The relationship between immigration and welfare support is at the heart of contemporary welfare politics literature. Recent research suggests that how Americans think about immigrants and immigration shapes their welfare attitudes (Haselswerdt 2020; Hussey and Pearson-Merkowitz 2012; Garand, Xu, and Davis Reference Garand, Xu and Davis2017; Myers, Zhirkov and Trujillo 2022), and the magnitude of this “immigration effect” even surpasses the effect of attitudes toward Blacks, leading to the argument that public opinion toward the contemporary American welfare state has been “immigrationalized” (Garand, Xu, and Davis Reference Garand, Xu and Davis2017).
Although scholars have connected immigration attitudes with public welfare support, it is not yet clear how information in individuals’ contexts and their immigration attitudes work in tandem to influence their welfare attitudes. Scholars have long documented the importance of “context” in shaping public opinion (Moore et al. 2003; Kimmel and Simone 2012; Hero and Tolbert 2004; Avery and Fine 2012; Xu and Garand Reference Xu and Garand2010), but the role of context in shaping the immigration-welfare linkage has not been fully explored. How individuals translate their immigration attitudes into support for/opposition against public welfare is an intriguing question.
Guided by comparative cross-national scholarship and state politics literature on the welfare state, we explore how immigration attitudes and immigrant welfare context combine to influence support for the welfare state. We argue that the immigration-welfare linkage is dependent upon contexts: citizens’ dislike or fear of immigrants will make them less supportive of public welfare, but this relationship is moderated by the levels of immigrant welfare participation rates in individuals’ contexts. Specifically, higher levels of immigrant welfare consumption in individuals’ contexts will likely amplify this effect and translate negative immigration attitudes into even lower welfare support.
To test this conditional effect of context on the immigration-welfare linkage, we merge contextual data on immigrant welfare participation rates and other control variables with public opinion data from the Cumulative American National Election Study (CANES) over the period of 2004–2016. Our estimates from the multilevel models reveal strong and consistent evidence that Americans’ immigration attitudes and the immigrant welfare participation rates in their home states combine to influence public welfare support. Individuals with favorable and unfavorable views of immigration respond very differently to higher levels of immigrant welfare participation, with the former translating high levels of immigrant welfare participation into more support for public welfare and the latter translating immigrant welfare participation into reduced support for public welfare. Further, the effects of Americans’ immigration attitudes on support for public welfare are considerably stronger for individuals who reside in states with high-welfare participation but are somewhat muted for individuals who reside in states with low-welfare participation.
This paper differs from previous empirical research by introducing the conditional effect of context on the relationship between immigrant attitudes and welfare support. Building upon previous literature on the “immigrationalization” of the American welfare state, our research suggests that public attitudes toward the contemporary American welfare state are influenced not only by Americans’ general evaluations of immigrants but also by the reality of welfare usage by immigrants in individuals’ contexts. These findings have important implications for the future development of the American welfare state.
Immigration Attitudes, Contexts, and Welfare Support
Scholars have long associated race with the development of the welfare state in the United States, and the bulk of this research has focused on race in terms of Blacks and Whites (Lieberman Reference Lieberman, Schram, Soss and Fording2003; Quadagno Reference Quadagno1994; Gilens Reference Gilens1996, Reference Gilens1999; Alesina and Glaeser Reference Alesina and Glaeser2004). Numerous studies have found that the size of the African American population in American communities, cities, and states is negatively related to the level of social welfare spending (Quadagno Reference Quadagno1994; Alesina, Baqir, and Easterly Reference Alesina, Baqir and Easterly1999; Luttmer Reference Luttmer2001; Alesina and Glaeser Reference Alesina and Glaeser2004). However, the changing demographics of the country caused by a large influx of immigrants in the last few decades suggest that a shift in focus may be necessary. Early work in the post-Aid to Families with Dependent Children (AFDC) environment suggests that policy choices and benefit levels are related to the size of states’ noncitizen populations (Hero & Preuhs, Reference Hero and Preuhs2007; Soss et al. Reference Soss, Schram, Vartanian and O’Brien2001; Xu Reference Xu2017). Using survey data from 1992 to 2012, Garand et al (Reference Garand, Xu and Davis2017) find that immigration attitudes are associated with individuals’ welfare attitudes, and the magnitude of this “immigration effect” is larger than the “race effect.” Haselwerdt (2021) further explores the causal mechanism of the immigration-welfare linkage in an experimental study and finds that respondents will link their immigration attitudes with support for welfare but only when they are primed and assume that immigrants are a “fiscal threat.”
Except for Haselwerdt’s (2021) experimental study, few empirical studies have attempted to test the causal mechanisms underlying the immigration-welfare linkage. We build on the work of Garand et al. (Reference Garand, Xu and Davis2017) and Haselwerdt (2021) to investigate how the immigration-welfare linkage is conditional upon contexts such as immigrant welfare consumption. We argue that individuals do not simply translate their perceptions of immigrants into lowered support for welfare; instead, they combine immigrant-related information in their environments with their existing immigration attitudes to update their support for the welfare state.
This interaction effect between immigration attitudes and immigrant welfare participation in context can be interpreted in two alternative ways: (1) negative immigration attitudes are associated with lowered support for public welfare, but this effect should be amplified among individuals living in states with high levels of immigrant welfare participation rates; (2) higher levels of immigrant welfare participation in individuals’ contexts will lead them to lower their support for public welfare, but this effect will be amplified among individuals with negative immigration attitudes. In the two sections below, we explain these two alternative interpretations of the same interaction effect that captures the interconnections among immigration attitudes, immigrant-related context, and public welfare support.
Immigration Attitudes and Welfare Support
Perhaps, the most common theoretical argument regarding immigration and welfare is that immigrants decrease the social solidarity that undergirds support for public goods. Because most immigrants entering the United States in the past few decades were Latino and Asian immigrants, their skin tones, language skills, and cultural backgrounds are different than those of the two largest racial groups, Whites and Blacks. In addition, immigrant populations often speak their native language or speak English with an accent, and it can take a long time for immigrants to be socially and culturally assimilated into the host society. As a result, immigrants often “look,” “act,” and “sound” different than native-born citizens and are often easily recognized as members of an out-group.
The “out-group” status of immigrants could affect American citizens’ general evaluations of immigrants as well as their perceptions of how deserving immigrants are of welfare benefits. It can also result in stereotypical images of the immigrant group that tend to be negative. Previous research suggests that the dominant stereotypes of immigrants is that immigrants are incompetent and untrustworthy (Cuddy et al. 2009; Lee and Fiske 2006). For many Americans, this stereotype of “incompetency and untrustworthiness” is the go-to image for immigrants, and they need additional information to be able to move away from the generic immigrant image (Lee and Fiske 2006). Other stereotypes of immigrants include “undeserving,” “fraudulent,” “irresponsible,” “negligent,” “noncontributing,” and “burden to taxpayers,” which were the terms commonly used in the congressional policy hearings leading to the 1996 welfare reform (Yoo 2008). In addition, poverty is more likely to be attributed to a lack of effort on the part of the individual for immigrants than for a generic poor person (Alesina et al. Reference Alesina, Miano and Stantcheva2023).
These negative stereotypes of immigrants can be directly linked to Americans’ perceptions of welfare deservingness. Scholars who have examined the rank ordering of deservingness among welfare recipients found that immigrants are perceived to be the least deserving, and this pattern is universal across individuals with different demographic backgrounds and across cultures and societies (Appelbaum 2002; van Oorschot Reference van Oorschot2006; 2008). Therefore, individuals’ attitudes toward immigrants (or immigration) are linked to their support for the public welfare system because citizens with negative opinions about immigrants are more likely to perceive them as members of an out-group, noncontributing, incompetent, untrustworthy, and a burden to taxpayers. These feelings will likely generate less support for the public welfare system, which is perceived as dispensing benefits to immigrant beneficiaries who are regarded as undeserving (Van Oorschot Reference van Oorschot2006; Burgoon and Rooduijn 2021).
Existing work on the social construction of target populations also suggests that how individual citizens think of policy beneficiaries will influence the politics surrounding that policy (Schneider and Ingram 1993). The public perceptions of the welfare target populations in the US had, for a long time, been Blacks (Gilens Reference Gilens1996). Gilens (1996; Reference Gilens1999) argues that the American welfare state is “racialized” because many White Americans hold negative stereotypes about the work ethics of African Americans, view African Americans as undeserving, and perceive that African Americans disproportionately participate in major welfare programs.
Yet, in recent years, immigrants have been increasingly considered as a perceived target population for social welfare programs, not only in the US but also in European countries (Oorschot 2006; Mau and Burkhardt Reference Mau and Burkhardt2009; Eger Reference Eger2010, Larsen Reference Larsen2011; Schmidt-Catran and Spies 2016; Alesina, Miano, and Stantcheva 2019; Haselwerdt 2021). When citizens perceive immigrants as a target population for social welfare, how they think of immigrants will shape how they evaluate the social welfare system. In particular, if certain individuals have negative stereotypes or perceive immigrants as undeserving, they will lower their support for public welfare. For instance, Fox (2004) finds that evaluations of Latinos could also affect Whites’ welfare attitudes. Both a large Latino population in individuals’ contexts and negative perceptions of Latinos can combine to lower their support for welfare spending.
More recently, an increasing number of studies have also connected attitudes toward immigrants with evaluations of the social welfare system in the US (Haselswerdt 2020; Hussey and Pearson-Merkowitz 2012; Garand, Xu, and Davis Reference Garand, Xu and Davis2017; Myers, Zhirkov and Trujillo 2022). Anti-immigrant attitudes are also connected to lower levels of support for redistribution in Denmark, France, Britain, and Sweden (Roemer and Van der Straeten 2005; 2006; Larsen Reference Larsen2011). More recent experimental work indicates that priming a respondent to think about immigration leads to less support for a whole host of social safety programs (Alesina et al. Reference Alesina, Miano and Stantcheva2023).
Based on these arguments and studies, we propose our first hypothesis on the relationship between immigration attitudes and support for welfare:
H1: Negative attitudes toward immigrants/immigration lead to lower levels of support for public welfare.
Immigrant Welfare Participation and Welfare Support
While immigration attitudes may have an important impact on individuals’ welfare attitudes, immigrant-related information in individuals’ contexts may also influence their welfare attitudes. Scholars have long documented the importance of “context” in shaping public opinion (Moore et al. 2003; Kimmel and Simone 2012; Hero and Tolbert 2004; Avery and Fine 2012). For example, as scholars examine the role of race in American politics, they suggest that one of the keys to unlocking its influence is “identifying the specific circumstances under which race is politically consequential” and that race is most likely to influence policy opinions when priming or framing are at work (Weber, Lavine, Huddy, and Federico Reference Weber, Lavine, Huddy and Federico2014, 63; Hopkins 2010; Huber and Lapinski 2006; Mendelberg 2008; Valentino, Hutchings, and White 2002; Weber et al Reference Weber, Lavine, Huddy and Federico2014.) Information in one’s context serves as an important priming factor, and the key to unlocking the linkage between immigration attitudes and public welfare support can be found in the types of contextual information with a priming effect (Alesina et al Reference Alesina, Miano and Stantcheva2023). Because immigrants are perceived as being among the most undeserving welfare recipients (Van Oorschot Reference van Oorschot2006; Burgoon and Rooduijn 2021), a high level of immigrant welfare participation in one’s context may serve as a priming factor and trigger Americans to lower their support for the welfare state.
Individuals could translate objective rates of immigrant welfare participation in their context directly into evaluations of the welfare system. For example, when individuals witness an immigrant with a strong accent using food stamps in the supermarket, or when they encounter an immigrant with a translator using Medicaid benefits in an emergency room in the hospital, they will be more likely to perceive that these immigrants abuse the welfare system designed for citizens and then subsequently further reduce their support for governmental welfare spending. Likewise, when individuals read or hear about stories of immigrants using various welfare benefits in their states in newspapers, on TV, or in their conversations with friends and family, they will also likely link the objective reality of immigrant welfare participation in their states to lowered support for the welfare system.
Since the 1996 Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA), states were granted discretion to fund immigrant welfare benefits using their own state funds (Zhu and Xu Reference Zhu and Xu2015; Xu and Zhu Reference Xu and Zhu2022). As a result, states adopted different immigrant welfare policies with varying levels of inclusiveness, which led to a wide range of immigrant welfare participation rates across the states. For example, from 2004 to 2016, the immigrant welfare participation rate in the 50 American states ranged from 11.7% to 60.0% with an overall average of 40.3%.
The objective immigrant welfare participation rates in one’s context are likely to be at least somewhat accurately perceived by citizens. Scholars find that individuals can have a reasonable and fair assessment of the reality of immigrant welfare participation rates in their home states (Xu, Garand, Davis, and Henderson Reference Xu, Garand, Davis and Henderson2023). When comparing individuals’ subjective perceptions of immigrant welfare participation with the actual objective immigrant welfare participation rates in their home states, Xu et al. (Reference Xu, Garand, Davis and Henderson2023) find a significant association between the perceived and actual immigrant welfare participation rates in individuals’ home states. These findings are supported by work from Alesina et al. (Reference Alesina, Miano and Stantcheva2023), who find that the actual economic circumstances of immigrants predict survey respondents’ perceptions of immigrants’ economic weakness.
Given the reasonably strong association between the actual immigrant welfare participation rates in one’s context and the perceptions of immigrant welfare participation, we have stronger reasons to believe that immigrants’ participation in welfare programs in individuals’ home states will likely prompt perceptions among some Americans that immigrants abuse the welfare resources designed for citizens and subsequently lower support for governmental welfare spending.
Therefore, we develop our second hypothesis and contend that variation in immigrants’ participation rates in social welfare programs across states is related to variation in individuals’ support for social welfare.
H2: Higher levels of immigrant welfare participation in individuals’ home contexts will decrease their support for public welfare spending.
Conditional Effects: Immigration Attitudes and Immigrant Welfare Participation
Individuals’ immigration attitudes and the immigrant welfare participation in their context not only directly influence support for welfare but could also have an interaction effect on welfare support. As Haselwerdt (2021) points out, the effect of immigration attitudes on social welfare support hinges upon the assumption that immigrants benefit from welfare. In an experimental study, Haselwerdt (2021) finds that when the respondent assumes that immigrants will benefit from welfare, their attitudes about immigration are a powerful predictor of welfare opposition; but when the respondent assumes that immigrants will not benefit, immigration attitudes are largely unrelated to welfare attitudes. Built on this finding, we argue that the relationship between immigration attitudes and welfare support is conditioned by how much immigrants actually benefit and participate in welfare programs. Alternatively, the relationship between immigrant welfare participation in context and welfare attitudes is conditioned by one’s immigration attitudes.
There are two alternative ways to explain the interaction effect. First, in Hypothesis 1, we have argued that individuals’ immigration attitudes affect welfare support because of negative stereotypes of immigrants and perceptions of immigrants as being undeserving of welfare benefits. Building on this hypothesis, we contend that levels of immigrant welfare participation rates in individuals’ environments could moderate the relationship between immigration attitudes and welfare attitudes. The degree to which individual citizens observe evidence that immigrants in their environment are using welfare benefits should serve as a priming factor that magnifies the relationship between immigration attitudes and support for the welfare state.
The logic behind this conditional effect is simple: suppose there are two groups of citizens, Group A and Group B, both of whom have negative attitudes toward immigrants. Group A lives in a state with very low levels of immigrant welfare participation, while Group B lives in a state with very high levels of welfare consumption by immigrants. Higher levels of immigrant welfare consumption in Group B’s environment will likely amplify their concern and translate their negative immigration attitudes into even lower welfare support. Yet for the members of Group A, low levels of immigrant welfare consumption in the environment may be less likely to prime them to translate the negative feelings about immigrants into decreased support for welfare. Based on this argument, we propose a conditional effect of the immigrant-related context on the relationship between immigration attitudes and welfare support, and that is: the effect of attitudes toward immigrants/immigration on welfare support should be amplified among individuals living in states with high levels of immigrant welfare participation rates.
This interaction effect can be interpreted in an alternative way. Based on Hypothesis 2, higher levels of immigrant welfare participation in individuals’ home contexts will decrease their support for public welfare. Built on this hypothesis, we argue that the relationship between immigrant welfare participation in context and one’s support for welfare can be conditioned by the individual’s immigration attitudes. In other words, even though immigrant welfare consumption in individuals’ home contexts could make them adjust their support for the public welfare system, how much individuals change their welfare attitudes in response to their home context should be conditioned upon how those individuals think of immigrants in the first place.
Attitudes toward immigrants provide a prism through which individuals evaluate and translate the immigration welfare rates in their home contexts into welfare attitudes. Individuals with negative opinions about immigration and immigrants will likely translate high levels of immigrants’ welfare consumption into decreased support for the welfare system. Yet, individuals with strong positive views toward immigrants will not necessarily lower their support for public welfare even when they see high levels of immigrant welfare consumption in their environment. Instead, they may show more sympathy and think that immigrants need more help. Therefore, we argue that the negative effect of immigrant welfare participation rates on support for public welfare spending will be amplified among individuals with negative opinions toward immigrants/immigration, but attenuated among individuals with positive opinions toward immigrants/immigration.
Based on these arguments, we propose an interaction effect between immigration attitudes and contextual immigrant welfare participation and argue that these two factors together shape one’s welfare support levels. We develop our third hypothesis based on this interaction effect:
H3: Individuals’ immigration attitudes and the immigrant welfare participation level in their context interactively shape public welfare support levels; individuals with negative immigration attitudes who live in high levels of immigration welfare participation will see the lowest levels of support for public welfare.
Data and methods
In order to explore the mechanism that explains the effects of immigration attitudes and immigrant welfare participation on public welfare support in the United States, we merge state-level contextual data on immigrant welfare participation rates with survey data from the CANES. Descriptions of key variables in the model can be found below, and in Appendix 1, we present detailed descriptions of the definitions, data sources, and descriptive statistics of these and other variables. Our analyses are conducted using data from four presidential election years—2004, 2008, 2012, and 2016.
Dependent Variable
Support for government welfare spending . Our dependent variable is an ordinal measure of support for welfare spending, coded 2 for respondents who support increases in welfare spending, 1 for those who would like to keep welfare spending at current levels, and 0 for those who prefer decreases in welfare spending. Because our dependent variable is an orderable discrete variable, and since we include both state- and individual-level independent variables in our models, we estimate our models using multi-level (hierarchical) ordered logit with random constants (Steenbergen and Jones Reference Steenbergen and Jones2002).
Key independent Variables
Attitudes toward immigration and immigrants. Following Garand et al. (Reference Garand, Xu and Davis2017), we use two questions from the CANES to capture individual attitudes toward immigration and immigrants. The first question asks about the degree to which individuals support increases in immigration; this variable ranges from 4 (for respondents who strongly support increased immigration) to 0 (for respondents who strongly support decreased immigration). The second question asks respondents to evaluate illegal immigrants using a feeling thermometer, which ranges from 0 (negative views toward illegal immigrants) to 100 (positive views toward illegal immigrants). We use principal components factor analysis to create a general measure of immigration attitudes based on these two variables, and our factor analysis shows that the two indicators load strongly on a single factor (Eigenvalue = 1.339, variance explained = 0.670). Based on our Hypothesis 1, we expect immigration attitudes to have a positive effect on welfare support.
Immigrants on welfare . In order to capture the reality of immigrants’ welfare participation at the state level, we use data from the Current Population Survey Annual Social and Economic Supplements (CPS-ASES) to tabulate the percentage of foreign-born households that participate in any major welfare programs in each state year.Footnote 1 Based on our Hypothesis 2, we expect that the coefficient for this variable will be negative, indicating that Americans residing in states with high levels of immigrant welfare participation will hold less favorable views toward welfare spending.
Immigration Attitudes × Immigrants on welfare: To capture the interaction effect of the two variables, we include the multiplicative term of the two variables in the model. We expect the coefficient for this variable to be positive, indicating that (1) the negative effect of immigrant welfare participation on support for welfare spending will be attenuated as attitudes toward immigrants become more favorable and (2) the positive effect of immigration attitudes on support for welfare spending will be amplified for individuals residing in states with high-immigrant welfare participation rates.
Control Variables
We include in our models a range of aggregate- and individual-level independent variables that serve as control variables. Specifically, at the aggregate level, we include measures of the immigrant share of the population, the Black population share, the share of Blacks on welfare, real state GDP per capita, the state unemployment rate, and state government ideology. For individual-level control variables, we include separate feeling thermometers for Blacks and for the poor, partisan identification, political ideology, and a series of demographic and socioeconomic variables (i.e., gender, race/ethnicity variables for Blacks, Hispanics, and Asians, age, education, family income, and frequency of church attendance). To maintain brevity, we include a detailed description of these variables in Appendix 1.
Model Estimation
For our individual-level analyses, the dependent variable—support for welfare spending—is an orderable discrete (ordinal) variable; so, we estimate our models using ordered logit. However, because we include contextual variables representing state immigrant welfare participation and other state-level variables, our data are clustered by state; with survey respondents nested within states, all respondents within specific states are assigned the same values on our state-level contextual variables. Given this, we estimate our models using multi-level (hierarchical) ordered logit (Gelman and Hill 2007; Long and Freese 2014).Footnote 2 ,Footnote 3
Empirical Results
To test our hypotheses, we estimate two models, with support for public welfare spending as our dependent variable in each model. The results for both models are reported in Table 1. As one can see, the first model permits us to estimate the direct effects of immigrant welfare participation rates in respondents’ home states and individuals’ immigration attitudes on Americans’ support for welfare spending. The second model tests the interaction effect for immigration attitudes and immigrant welfare participation rates, suggesting a significant interaction effect of these two variables on welfare support. In order to facilitate the interpretation of interaction effects, we have rescaled the variables measuring the state-level immigration welfare participation rates and immigration attitudes to range in value from 0 to 1. The coefficients for these variables can be interpreted to represent shifts from the minimum to maximum values on these variables.
Table 1. Multi-level ordered logit model estimates for support for public welfare spending in the United States, 2004–2016
*** prob < 0.001 ** prob < 0.01 * prob < 0.05.
Direct effects of immigration attitudes and immigrant welfare participation
In Model (1), we report the results of a hierarchical ordered logit model that includes immigrant welfare participation rates and immigration attitudes, along with control variables but without the interaction variable. We observe that immigration attitudes have a strong positive effect on welfare support (b = 1.842, z = 16.70), suggesting that individuals with more positive views about immigrants are substantially more likely to support welfare spending. As shown in Figure 1, moving from the lowest to the highest value of immigration attitudes (and holding other variables constant at their means) results in a downward shift in predicted probabilities for support for decreased welfare spending of −0.398 (from 0.586 to 0.188), an upward shift in support for increased welfare spending of 0.233 (from 0.064 to 0.297), and an upward shift for keeping welfare spending the same of 0.165 (from 0.350 to 0.515). Clearly, as favorability toward immigrants increases, respondents are significantly less likely to support decreases in welfare spending and significantly more likely to support increases in welfare spending or to maintain welfare spending levels at current levels. This supports our Hypothesis 1 and confirms the findings by Garand et al. (Reference Garand, Xu and Davis2017) on the positive relationship between immigration attitudes and support for welfare spending.
Figure 1. Direct effect of immigration attitudes on support for welfare spending, 2004–2016 Cumulative American National Election Study.
We also find that immigrant welfare participation rates have a significant negative direct effect on public welfare support (b = −0.634, z = −2.41), which confirms our Hypothesis 2. As shown in Figure 2, as one moves from the lowest to highest level of immigrant welfare participation rates, there is a discernible increase in support for decreased welfare spending of 0.146 (from 0.313 to 0.459) and a decrease in support for increasing welfare spending of −0.074 (from 0.177 to 0.103) and of keeping welfare spending the same of –0.073 (from 0.511 to 0.438).
Figure 2. Direct effect of state immigrant-welfare participate rate on support for welfare spending, 2004–2016 Cumulative American National Election Study.
Overall, Model (1) makes clear that there is a strong positive relationship between Americans’ immigration attitudes and welfare support. Moreover, there is also a significant negative relationship between immigrant welfare participation and welfare support. These findings are consistent with our Hypotheses 1 and 2.
Interaction effects for immigration attitudes and immigrant welfare participation
In Model (2), we consider the joint effects of immigration attitudes and state-level immigrant welfare participation rates by including the interaction variable for these two variables. Our results suggest that each of these variables moderates the effect of the other on support for welfare spending. Turning first to immigration attitudes, we find that the coefficient for this variable is, as expected, positive and statistically significant (b = 0.916, z = 2.05); this represents the conditional effect of immigration attitudes on support for welfare spending, calculated when immigrant welfare participation rate is set equal to its lowest value (i.e., 0 in the rescaled variable). Even in those states with low-immigrant welfare participation rates, immigration attitudes have a significant (albeit moderate) positive effect on welfare support, indicating that those with favorable (unfavorable) attitudes toward immigrants are more (less) supportive of welfare spending. Second, we find that immigrant welfare participation has a significant negative effect on support for welfare spending (b = −1.229, z = −3.22). The coefficient represents the effect of this variable among respondents who have the most unfavorable views toward immigration (i.e., when immigration attitudes are set equal to 0 on the rescaled variable). As we expect, among individuals who have negative views toward immigration, there is a strong negative effect of immigrant welfare participation rates on welfare spending attitudes.
The main storyline here is found in the coefficient for the interaction variable for immigrant welfare participation and attitudes toward immigration. This coefficient is positive and statistically significant (b = 1.451, z = 2.14), suggesting a strong positive interaction effect of these two variables on welfare support. This interaction coefficient indicates that immigration attitudes and immigrant welfare participation have a moderating effect on each other as each variable shapes welfare spending attitudes. This finding confirms our Hypothesis 3.
These results demonstrate that the effect of immigrant welfare participation rates on welfare spending attitudes increases (and becomes less negative) as individuals move from unfavorable attitudes toward immigration to favorable attitudes toward immigration. We have already noted that there is a negative effect of immigrant welfare participation for individuals with the most unfavorable immigration attitudes (b = –1.229), but for individuals with the most favorable immigration attitudes, the negative effect of immigrant welfare participation shifts to the positive (b = +0.222, or −1.229 + [1 * 1.451]). It is important to note that at this level of immigration attitudes, the positive effect of immigrant welfare participation on support for welfare spending is statistically indistinguishable from 0 (χ2 = 0.21, prob > 0.6434).
In addition, the interaction coefficient tells us how the effect of immigration attitudes on welfare spending attitudes becomes stronger and more positive as immigrant welfare participation shifts from its lowest to highest values. For individuals residing in states with the lowest immigrant welfare rate, there is a positive effect of immigration attitudes on support for welfare spending, as already noted (b = 0.916), but for individuals from states with the highest immigrant welfare rates, the effect is even stronger (b = 2.367, or 0.916 + [1 * 1.451]) and highly significant (χ2 = 77.35, prob > 0.0000). The bottom line is that there is strong evidence that (1) immigration attitudes moderate the effect of immigrant welfare participation on support for welfare spending and (2) immigrant welfare participation moderates the effect of immigration attitudes on support for welfare spending.
Because the coefficients in an interaction model with a three-point ordinal dependent variable are difficult to interpret, we present figures that demonstrate how predicted probabilities on the dependent variable change as a function of different combinations of immigration attitudes and immigrant welfare participation.
Immigration attitudes moderate effects of immigrant welfare participation. We suggest that the relationship between immigrant welfare participation and support for welfare spending is dependent on the degree to which individuals have favorable immigration attitudes. In Figures 3a, 3b, and 3c, we show that there are major differences in the effects of immigrant welfare participation on support for different outcomes on the welfare spending variable for those with favorable and unfavorable immigration attitudes. To begin, in Figure 3a, we generate predicted probabilities for support for decreased welfare spending, estimated separately for those with positive and negative immigration attitudes. As one can see, increases in immigrant welfare participation yield increases in the probability of support for decreasing welfare spending among individuals with negative immigration attitudes. The increase is from 0.396 (for those residing in states with the lowest level of immigrant welfare participation) to 0.687 (for those who reside in states with the highest level); this is an increase of 0.297 across the range of the immigrant welfare participation variable. This is what we would expect: individuals with negative immigration attitudes should become more likely to support decreased welfare spending as immigrant welfare participation in their home state rises to its highest level. For individuals with positive views toward immigration, increases in immigrant welfare participation in their contexts only slightly depress support for decreased spending on welfare programs; the effect is relatively small (only from 0.210 to 0.176, a difference of 0.034) for those with positive immigration attitudes.
Figure 3. Predicted probabilities for effects of immigrant-welfare participation on support for increased welfare spending, keeping welfare spending the same, or decreased welfare spending, for negative and positive immigration attitudes, 2004–2016 American National Election Study.
What is the effect of immigrant welfare participation on support for keeping welfare spending at current levels? In Figure 3b, we present the predicted probabilities across the range of immigrant welfare participation rates for individuals with both positive and negative immigration attitudes. For individuals with negative immigration attitudes, increases in immigrant welfare participation results in a discernible decline in support for maintaining current welfare spending levels; the decline is from 0.475 for those with the lowest immigrant welfare participation rate to 0.271 for those with the highest rate—a shift of 0.204 across the range of this variable. On the other hand, for individuals with positive immigration attitudes, there is almost no effect of the immigrant welfare rate on support for keeping welfare spending the same; the shift across the range of immigrant welfare participation is only −0.011 (from 0.521 to 0.510). Clearly, for those with negative immigration attitudes, increases in immigrant welfare participation depresses support for keeping welfare spending at current levels. However, for those with positive immigration attitudes, the level of immigrant welfare participation does not have a significant effect on welfare support.
When it comes to support for increased welfare spending (Figure 3c), we see that the curves for those with positive and negative immigration attitudes are not too far apart at low levels of immigrant welfare use, but as immigrant welfare use increases, the two curves diverge, with individuals holding positive immigration attitudes increasing their probability of supporting increases in welfare spending and respondents holding negative immigration attitudes decreasing their probability of supporting welfare spending increases. Specifically, at low levels of immigrant welfare participation, individuals with positive and negative immigration attitudes differ in their probability of support for increased welfare spending by about 0.140 (i.e., 0.269–0.129), but as immigrant welfare participation increases the gap in the probability of supporting increased welfare spending rises to almost 0.272 (i.e., 0.314–0.042). This represents an increased gap of 0.132 between the two curves over the range of immigrant welfare participation rates. Simply, the effect of immigrant welfare participation on support for increased welfare is significantly different for those with positive and negative immigration attitudes.
Immigrant welfare participation moderates the effects of immigration attitudes. Strong theoretical foundations also support the argument that immigrant welfare participation moderates the relationship between immigration attitudes and welfare support. In Figures 4a, 4b, and 4c, we present (respectively) predicted probabilities of support for decreased, maintained, and increased welfare spending as a function of immigration attitudes, broken down for individuals from states with low and high rates of immigrant welfare participation.
Figure 4. Predicted probabilities for relationship between immigration attitudes and support for increased welfare spending, keeping welfare spending the same, or decreased welfare spending, for low- and high-immigrant-welfare participation rates, 2004–2016 American National Election Study.
Turning first to support for decreased welfare spending (Figure 4a), we show that the relationship between immigration attitudes and welfare attitudes differs considerably depending on the immigrant welfare participation rates in individuals’ home states. For respondents who live in states with low-immigrant welfare participation, there is a moderate negative relationship between immigration attitudes and support for decreased welfare spending; moving from the least favorable to most favorable immigration attitudes results in a shift in the probability of support for decreased welfare spending of 0.186 (i.e., 0.396–0.210). On the other hand, the effect of immigration attitudes is, as expected, much stronger for individuals who reside in states with high-immigrant welfare participation. Moving from the lowest to the highest level of immigration attitudes, there is a shift in the probability of supporting a decreased welfare spending of 0.511 (i.e., 0.687–0.176). This suggests a substantial difference of 0.325 (i.e., 0.511–0.186) in the effects of immigration attitudes on welfare support in states with low- and high-welfare participation rates. Clearly, the effects of immigration attitudes on support for decreased welfare spending are magnified for respondents living in states with high-immigrant welfare participation rates.
For individuals’ support for keeping welfare spending at current levels, we also find a discernible difference between respondents residing in states with low- and high-immigrant welfare participation (Figure 4b). For those living in states with low levels of immigrant welfare participation rates, there is only a weak effect of immigration attitudes on support for keeping welfare spending the same; as one moves from the most unfavorable to the most favorable immigration attitudes, we find a shift in the probability of support for keeping welfare spending at current levels of only 0.046 (i.e., 0.521–0.475). On the other hand, for respondents who reside in states with high-immigrant welfare participation, there is a much larger shift in support for keeping welfare spending at current levels of 0.239 (i.e., 0.510–0.271). This represents a difference of 0.193 (i.e., 0.239–0.046) in the effects of immigration attitudes on keeping welfare spending the same for those in states with low and high levels of immigrant welfare participation.
Finally, what about the effects of immigration attitudes on support for increases in welfare spending? In Figure 4c, we present predicted probabilities of support for increased welfare spending as a function of immigration attitudes, shown separately for individuals residing in states with low- and high-immigrant welfare participation rates. Here again, for those in low-immigrant welfare participation states, there is a moderate effect of immigration attitudes on support for increased welfare spending; moving from the lowest to the highest value on immigration attitudes, there is a probability change of 0.140 (0.269–0.129), suggesting a discernible effect of immigration attitudes on the probability that individuals support increased welfare spending. But for those living in states with high-immigrant-welfare participation, the effects of immigration attitudes are considerably larger: moving from the lowest to highest value on immigration attitudes in states with high-immigrant-welfare participation results in a shift in support for increased welfare spending of 0.272 (i.e., from 0.042 to 0.314), an effect that is almost double the effect for respondents residing in states with low-immigrant-welfare participation. It is clear that the effects of immigration attitudes on support for increased welfare spending are stronger for individuals who reside in states with heightened immigrant presence in the welfare system than in states with a low-immigrant presence.
Overall, our findings suggest that for those living in states with low-immigrant welfare participation, immigration attitudes do not have a particularly strong effect on support for welfare spending. When there are relatively few immigrants who are recipients of welfare programs, holding favorable or unfavorable views toward immigration does not have a large effect on shaping how individuals think about welfare spending. But for those living in states with high-immigrant welfare participation, the effect of immigration attitudes on support for welfare spending is discernibly stronger, suggesting that the presence of a larger immigrant welfare population in one’s context has the effect of activating the linkage between how one thinks about immigration and how one thinks about welfare.Footnote 4
Taken as a whole, our empirical results suggest that Americans’ support for welfare spending is a function of a combination of attitudes toward immigration and the immigration welfare participation rate in their home state contexts. These two variables work together to determine whether individuals support higher, lower, or constant levels of welfare spending. We suggest that Americans’ immigration attitudes act as a prism that determines how immigrant welfare participation rates are interpreted and translated into welfare attitudes. Americans with negative immigration attitudes respond negatively to higher rates of immigrant welfare participation, while Americans with positive immigration attitudes respond either neutrally or moderately positively. Hence, the relationship between immigrant welfare participation and support for welfare spending is conditioned on the favorability or unfavorability of Americans’ immigration attitudes.
But it is also the case that the relationship between immigration attitudes and support for welfare spending is moderated by the immigration welfare participation rates in one’s context. When immigrant welfare participation is low, the contextual signals associating immigrants with welfare use are diminished. As a result, individuals are less likely to strongly connect their immigration attitudes with their support for welfare spending. High-immigration welfare participation can activate that linkage between immigration attitudes and support for welfare spending, in the sense that high-immigrant welfare participation provides a contextual framing to support the perception of a relationship between immigration attitudes and welfare spending attitudes. Simply, if there are few immigrant welfare participants in individuals’ contexts, there is reduced reason to connect immigrants and welfare; in contrast, if there are many immigrant welfare participants in an individual’s context, the connection between immigration and welfare becomes clearer and stronger.
It’s important to note that while attitudes toward immigrants influence attitudes toward welfare spending, measures assessing attitudes toward Blacks, Black participation in the social safety net, and Black population size fail to reach statistical significance. In addition, the coefficients for the percentage of a state’s foreign-born population are positive and significant. These findings provide evidence for the claim that welfare has become immigrationalized.
Robustness tests
Even though our analyses show strong support for our hypotheses, several important issues need to be considered. First, what conclusions can be reached from simpler model specifications that do not include the full range of control variables that we report in our results? In Appendix Table 2, we consider both the direct and interaction effects of state immigrant welfare participation and immigration attitudes on support for welfare spending, estimated without our wide set of control variables. As one can see, our results based on models without control variables are similar to those based on models that include control variables. In each case, we find a strong positive effect of immigration attitudes on support for welfare spending, as well as strong negative effects of immigrant welfare participation; this is the case for specifications that both exclude and include state-fixed effects.
Second, we consider the effects of an alternative measure of immigrant welfare participation. Our core measure represents immigrant welfare households as a percentage of all immigrant households, but an alternative measure represents immigrant welfare households as a percentage of all welfare households. In Appendix Table 3, we present results that are based on substituting the alternative measure of immigrant welfare participation for our original measure. As one can readily see, the results for Table 1 and Appendix Table 3 are strikingly similar. In both cases, immigration attitudes have a strong positive effect, and immigrant welfare participation rates have a strong negative effect on support for welfare spending (Models 1 and 2), and in both cases, the interaction coefficient is positive and statistically significant. Therefore, our two key independent variables have both direct and interaction effects on support for welfare spending, regardless of which measure of immigrant welfare participation we use.
Finally, it is possible that the relationship between immigration attitudes and welfare attitudes is endogenous; in other words, individuals’ welfare attitudes could also influence their immigration attitudes. To address this, we use a two-stage least-squares model with instrumental variables (IV-2SLS) to ensure that the effect of our independent variable (i.e., immigration attitudes) on our dependent variable (i.e., welfare attitudes) still holds after we “purge” the endogenous component of the independent variable. Using the 2016 ANES data as an example, we estimate an IV-2SLS model, and the results show strong support for a causal arrow that leads from immigration attitudes to welfare attitudes even after we consider possible endogeneity. The results of this endogeneity test are found in Appendix 4. In addition, the experimental studies by Alesina et al (Reference Alesina, Miano and Stantcheva2023) and Haselwerdt (2021) provide further evidence of the causal relationship between immigration attitudes and support for welfare, complementing our endogeneity test.
Summary and Conclusions
There is a growing body of evidence illustrating that how Americans think about immigration is strongly related to how they think about the welfare state. In this paper, we build on this body of research to explore further the mechanism through which immigration affects welfare support in the United States. Specifically, we suggest a process that combines immigration attitudes and the propensity of immigrants to be recipients of welfare programs (broadly defined) in individuals’ home state contexts in explaining support for welfare spending. Not only should immigration attitudes have a positive effect on support for welfare spending, but we also suggest that if Americans perceive that immigrants have a high propensity to be in the welfare system, they should be less likely to exhibit high levels of support for welfare spending. We test our arguments with ANES survey data for presidential election years from 2004 to 2016. Our results show strong evidence for the joint effects of immigration attitudes and immigrant welfare participation rates on attitudinal support for welfare spending.
Our results suggest that the effects of immigration attitudes and immigrant welfare participation rates on welfare support are conditioned by each other. Immigrant-welfare participation rates have a significant and negative effect on welfare attitudes, and this negative effect is magnified for those with strong negative immigration attitudes and becomes considerably weaker for those holding positive immigration attitudes. For individuals with negative immigration attitudes, the combination of negative views toward immigration and a high share of immigrants receiving welfare benefits in individuals’ contexts activates concerns about government services going to the “undeserving,” and the result is a decrease in support for welfare spending. However, when individuals hold positive immigration attitudes, high-immigrant welfare participation rates hold less power in swaying individuals to oppose welfare spending.
Our results can also be interpreted to say that state immigrant welfare participation moderates the effects of immigration attitudes on support for welfare spending. When individuals reside in states with low-immigrant welfare participation, there are diminished reasons for them to connect their immigration attitudes and support for welfare sending. On the other hand, a high-immigrant welfare participation rate in an individual’s home state signals a connection between immigration status and welfare participation. As a result, individuals are more likely to translate their immigration attitudes into either increased or decreased support for welfare spending.
A central contribution of this study is its examination of the role of context in shaping American public opinion at the intersection of immigration and welfare policy. Scholars have long recognized the importance of contextual factors in public opinion formation (Moore et al. 2003; Kimmel and Simone 2012; Hero and Tolbert 2004; Avery and Fine 2012; Marquart-Pyatt 2012). Prior research has established a relationship between objective contextual factors and citizens’ subjective perceptions of such contexts (Newman et al. 2013; Xu et al. Reference Xu, Garand, Davis and Henderson2023), yet the specific ways in which contextual factors influence welfare attitudes remain underexplored. Our study advances this literature by demonstrating both the direct and conditional effects of context on Americans’ welfare attitudes. Notably, we find that the public does not indiscriminately translate anti-immigrant sentiment into opposition to the welfare state. Rather, individuals incorporate contextual cues—such as the level of immigrant participation in welfare programs—into their evaluations, suggesting a more systematic and information-driven process. These findings challenge conventional assumptions about the dominance of affective heuristics in public opinion and highlight the substantive role of contextual information in shaping policy preferences. By elucidating the mechanisms through which context informs attitudes toward redistribution and immigration, this study contributes to the broader literature on public opinion formation and the intersection of welfare and immigration.
While further verifying the possible “immigrationalization” of Americans’ welfare attitudes, this research has important implications for the future development of the American welfare state. There has been considerable debate about the degree to which American immigration policy should focus on targeting high-skill immigrants who are likely to garner higher incomes and who are less likely to be part of the welfare system. Our findings suggest that immigration attitudes and the share of immigrants in the welfare system combine to shape Americans’ views toward welfare spending so changes in either of these variables could result in changes in welfare spending attitudes. A change in the mix of immigrants coming into the United States that favors high-skill immigrants (who will be less likely to be part of the welfare system) may have the effect of lowering the share of immigrants receiving welfare and hence may reduce anti-welfare intensity among those with negative views toward immigrants. On the other hand, if low-skill immigration continues to increase and the share of immigrants in the welfare system increases, support for the public welfare safety net may well continue to erode, particularly among those who hold less favorable views toward immigrants. The admonition that we can have (immigration-driven) diversity or welfare but not both seems to have some validity (The Economist, 2006).
These findings have important implications for the political rhetoric surrounding immigration, particularly in the current political climate. Our results indicate that immigrant welfare participation has the strongest impact on welfare attitudes among those who already hold negative views toward immigrants, while those with more positive attitudes remain largely unaffected. This suggests that if the Republican Party continues to emphasize anti-immigration rhetoric that explicitly ties immigration to welfare use, it could further reinforce the connection between negative immigration attitudes and opposition to welfare spending. During his term as president, Donald Trump frequently attacked immigration, often portraying immigrants as a burden on the welfare system. He reinforced this narrative through policy measures, including making welfare usage a barrier to green card eligibility. More recently, the 2024 presidential race has seen similar rhetoric, such as the focus on legal Haitian immigrants in Springfield, Ohio, where media coverage has linked crime allegations with increased participation in programs like Medicaid and SNAP (Reuters, 2024). These narratives, particularly when they evoke economic anxieties and welfare-related group cues (Brady et al. 2008), could not only deepen the relationship between anti-immigration sentiment and welfare opposition but also contribute to a broader erosion of public support for immigration itself.
Our research has several limitations. First, we use state-level immigrant welfare participation rates to capture contextual effects, and we realize that the contextual effects may very well exist at the city or county level instead. Therefore, our test based on states as the appropriate contextual level is a much more conservative test of our hypotheses, since we are looking at these patterns in a much broader and less localized context. The fact that we still find significant contextual effects suggests that our hypotheses will likely withstand tests with better and more localized contextual measures. In future studies, scholars should consider how contextual effects work at a more localized level (such as city, Zip Code, or county). Second, we use objective immigrant welfare participation rates to capture the effects of immigrant welfare threat because data on subjective perceptions of immigrant welfare participation are not available, at least not in the ANES data used here. We realize that there might be a gap between objective immigrant welfare rates and subjective perceptions of those rates, although we note that Xu et al. (Reference Xu, Garand, Davis and Henderson2023) do present evidence that these two variables are correlated. Third, we are unable to delineate the effects of various immigrant subgroups such as Latinos, Asians, Europeans, and Middle Easterners, as our study is only a first step in disentangling the causal mechanism between the immigrant-welfare attitudes linkage. In future research, it is important to explore further which subgroup(s) of immigrants have a stronger effect on Americans’ welfare attitudes. Last but not least, we have chosen to further explore the mechanism underneath the immigration-welfare linkage but have largely put aside the theoretical debates on the credibility of immigration-welfare causality, that is, whether or not the linkage is a credible causal relationship, or are both factors shaped by a general trend of social conservatism, as pointed out by Levy (2021). Future studies should look into this theoretical debate further.
The research program on the linkage between immigration and the welfare state is a full one. So far, we know very little about Americans’ perceptions of the immigrationalization of welfare. At the simplest level, to what extent do Americans perceive immigrants as high users of welfare programs? How do Americans perceive the relative level of welfare use for immigrants in comparison to other groups, including Blacks, Whites, native-born Hispanics, native-born Asians, and other groups? Further, what explains individuals’ perceptions of welfare usage rates among immigrants? Is it, as Gilens (Reference Gilens1999) suggests for Blacks, how the media covers immigrants in the United States? Is it that media coverage of welfare recipients includes a disproportionate number of immigrants? Alternatively, are Americans’ perceptions of welfare usage rates among immigrants a function of personal experience, uniformed stereotypes, or word of mouth through personal networks? We need to learn more about the degree to which Americans perceive immigrants as high users of welfare programs and the determinants of those perceptions. We encourage future research to explore these questions and further increase our understanding of the immigrant-welfare linkage in the United States.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/rep.2025.10012.
Funding statement
The authors of this paper did not receive any funding to support this project.
Open access
