Geographic Identity and Attitudes toward Undocumented Immigrants

Making a Place for Geographic Identity and Attitudes toward Immigrants and Immigration

A dominant strand of research on attitudes toward immigrants and immigration focuses on redistributive or economic threats to native workers or the so-called labor and economic competition model. There are mixed results on the extent to which immigrants, specifically undocumented immigration, affect natives’ wages. This model predicts that those with higher education and income will be more receptive to low-skill immigrants, than their counterparts, because they do not foresee competing with them (Fennelly and Federico 2008; Mayda 2006). Educational attainment is among the most consistently used predictors of both racial and policy attitudes. Well-educated whites tend to be more racially liberal, in that they support egalitarian ideals more so than the less educated (Federico 2004). Espenshade and Calhoun (1993) found that respondents with higher levels of educational attainment have more favorable attitudes toward undocumented immigrants. Therefore, we expect increased levels of education might lead to more progressive policy views toward undocumented immigration. However, those with less educational attainment may be more opposed to undocumented immigration.

Higher income residents may be less likely to live in the same areas as the undocumented—who tend to have lower incomes— further reducing the likelihood of feeling threatened. According to the competition hypothesis, we expect that respondents having high incomes will hold less punitive policy perspectives toward undocumented immigration because they are not in direct economic competition. Conversely, we expect that respondents with lower incomes may be less likely to support “open-door” policies that would allow undocumented workers to remain in the United States (Diaz, Saenz, and Kwan 2011).

Existing literature also shows support for the role of political predispositions, specifically ideological conservatism in negatively shaping attitudes toward undocumented immigration (Chandler and Tsai 2001; Espenshade and Hempstead 1996; Fennelly and Federico 2008). Several additional individual-level factors have been shown to influence policy attitudes toward immigration including demographic characteristics such as age (Citrin et al. 1997; Espenshade and Calhoun 1993), gender (Amuedo-Dorantes and Puttitanun 2011; Hood and Morris 1997), income and education (Federico 2004; Glaser 2001), local context (Ha 2010), as well as issue frames and news media coverage (Haynes et al. 2016, Merolla et al. 2013).

The role of racial/ethnic identity specifically regarding undocumented immigration reform remain understudied (but see Carter and King-Meadows 2019; Frasure-Yokley and Greene 2013; Morris 2000; Nteta 2013, 2014; Smith 2017; Smith and Greer 2018). African Americans, for example, have been relatively positive toward immigration when compared with other racial groups (Jones 2014). This is despite both mainstream media and research studies on the purported negative effects of immigrants on the employment opportunities of working class blacks (Borjas 1999, 2000; Burns and Gimpel 2000; Marrow 2011; Waldinger 1996). Even fewer studies have examined the role of racial identity specifically on attitudes toward undocumented immigrants among Asian Americans and Pacific Islander populations (Leung et al. 2019).

Existing work has focused on certain aspects of one’s environmental context such as the change in growth of the Latino population (Hopkins 2010; Johnston, Newman, and Velez 2015; Newman 2013; Rocha et al. 2011; Winders 2005) and the role of politicizing immigration population changes in the context of Latino population growth (Hopkins 2010). On one hand, some scholars find that closer contact with out-groups fosters greater understanding and empathy (Oliver 2010; Oliver and Wong 2003). According to Oliver and Wong (2003), this usually takes place when interaction between groups is facilitated near one another. To measure this, they focus on group attitudes given racial and ethnic composition at the census tract level. The intergroup contact literature suggests that under the right conditions, increased exposure and contact to out-groups will result in more positive racial attitudes (Cain, Citrin, and Wong 2000; Hood III and Morris 1997; Pettigrew and Tropp 2006). On the other hand, close proximity to out-groups can lead to increased hostility or racial threat (Enos 2016; Ha 2010; Key 1949; Reny and Newman 2018). These literatures have struggled to tease apart the true effect of context on one’s attitudes toward out-group members. Oliver and Wong (2003) showed that geographic units are key for understanding this relationship. At the census tract level, sizeable out-groups correlate with positive attitudes toward the group. However, at the county level, the larger out-groups correlated with more racially conservative attitudes.

What this line of research makes clear is the role and relevance of geography in structuring the types of interactions that occur among in-group and out-group members—which may influence how individuals develop out-group attitudes. However, the existing work has not sufficiently examined the variation in type of geography, but rather focused on the proximity to out-groups as the key variable of interest to understand attitudes toward out-group members. Contextualizing the impact of increased migration of blacks to California during the second great migration, Reny and Newman (2018) found that a sense of racial threat influenced white support for Proposition 14 in 1964. This ballot proposition would have allowed homeowners and landlords to discriminate on the basis of race. They found that proximity to growing black communities was associated with greater white support for the proposition.

In this article, we advance an argument that the type of place matters and that geographic types should be structured beyond the rural/urban binary or the rural/suburban/urban distinction. Even work that has split areas into rural/suburban/urban types misses the vast heterogeneity that exists within the United States’ suburban landscape. We argue that these community types not only differ along the rural–urban spectrum but also with respect to racial and ethnic characteristics, the tax capacity of the area, population density, home values, and so on.

We expect that areas farther away from the central city will be less supportive of undocumented immigrants as these are the areas that are likely to be least proximate to immigrant population (Oliver 2003). The racial and ethnic diversity within the central city and in the suburbs closest to the central city promote favorable attitudes toward out-groups giving the positive opportunities for contact in these locales.

Research Design, Data, and Methods

We use data from the 2016 CMPS, a multi-racial/ethnic, multi-lingual, post-election online survey in the United States (Barreto et al. 2017). The CMPS queried 10,145 people in five languages—English, Spanish, Chinese, Korean and Vietnamese. It includes large and generalizable samples of blacks (n = 3,102), Latinos (n = 3,003), Asian Americans (n = 3,006), and whites (n = 1,034), which allow for within-group comparison and analysis of an individual racial group, or comparative analysis across groups. ¹ We then merged the data with 2000 and 2010 decennial census and the 2015 American Community Survey (five-year estimates) for demographic data at the zipcode level. For a comprehensive description of the 2016 survey design and methodology, see Barreto et al. 2018.

Dependent Variable

We consider attitudes toward immigrants with the following question: “Which comes closest to your view about [SPLIT A: undocumented | B: illegal] immigrants who are already living and working in the U.S.?” This question was asked across all samples with the possible responses:

They should be allowed to stay in their jobs and apply for U.S. citizenship; (3)

They should be allowed to stay in their jobs, but temporarily; (2)

They should be required to leave their jobs and immediately leave the United States. (1)

We coded responses in an ordered fashion where higher values are indicative of more positive feelings toward immigrants. The embedded framing experiment is useful and ensures that our results are not driven by the terminology around immigrants but rather a respondent’s feelings toward immigrants (Figure 1). ²

OPEN IN VIEWER

Figure 1.

Distribution of the dependent variable.

Adaptation of Orfield’s Typology Using Aggregate-Level Census Data

The focus of this paper is to understand the differences in attitudes toward undocumented immigrants across various types of American urban, suburban, and rural areas and to examine more nuanced variations within these categories. While existing work has considered local context, it has often looked at factors such as population changes or binary classifications of rural/urban or metro/non-metro divisions. We seek to move beyond these factors and use a distinction of geographic types that better contextualizes the role of place (Lichter and Ziliak 2017; Scala and Johnson 2017).

American metros and rural areas have become more difficult to distinctly define along an urban/suburban/rural typology (Jackson 1985; Oliver 2001, 2003; Orfield 2002). For example, individuals are traditionally characterized as living in a suburb if they reside in the census-defined portion of a metropolitan area outside of a central city. The traditional definition fails to distinctly identify geographic areas when considering population characteristics such as median household income, household size, population density, growth rate, and employment opportunities, among many others (Cooke 2010; Garreau 1991; Lucy and Phillips 2000; Orfield 2002). Using the urban/suburban/rural typology does not allow for the heterogeneity in many American communities since its dividing lines are quite rigid. Following the rapid suburban expansion in World War II (WWII) and the more recent suburban and exurban changes, we need a classification schema that can help scholars tease apart these areas. In other words, the existing schema that have been used to classify the type of place have not been updated to reflect more recent suburban and exurban changes. However, academics and policy makers continue to rely on these longstanding classifications, which may not capture a true reflection of place-based distinctions.

In order to better understand the heterogeneity across geographic and spatial patterning, we map a more detailed taxonomy onto the respondents in a large public opinion survey. This taxonomy is based on an adaption of classifications originally defined by Orfield (2002) in his examination of the twenty-five largest metropolitan regions in the United States. In our study, Orfield’s classifications are applied at the zipcode level—rather than at the city or jurisdictional level—for a finer measurement scale that provides a more nuanced understanding of geographic variation and one that is meaningful for respondents’ understanding of place (Velez and Wong 2017). We suggest that this mapping is better able to understand the intricacies of place and how the politics of place relates to a variety of attitudinal outcomes, including those related to immigrants since it provides much more nuance and variation that we would obtain using county, region, metropolitan statistical area, or city.

Orfield’s original analysis included measures of fiscal characteristics, such as revenue capacity and expenditure need, as well as sociopolitical environmental factors as determinants of geographic type. This includes tax capacity and service environment cost measures, as well as a measure of racial and ethnic context. We use this typology since it provides a more refined way to understand some of the variation that exists across the metropolitan space along dimensions that relate to political outcomes. ³ In an attempt to follow the measures of Orfield, we constructed a classification schema where we use variables at the zipcode level to classify and detect the variation in geographic types across the United States’ spatial landscape. We include change in population growth (2000–2010), 2010 population density, and 2010 age of housing stock as measures of service environment at the local level. We include the percentage of those living in poverty in 2010, percentage unemployed in 2010, and the 2010 housing unit density as additional cost-based measures. The measures provide a look at place-based needs and levels of inequality in various locals. Tax capacity was measured using 2009 median household rent, 2009 median household value, and 2009 median household income. Notably, tax capacity is an important indicator of how high tax rates must be to support a given level of public services. The extant literature on the political economy of place contends that both residents and the business community want the highest value of goods and services for their tax dollar (Orfield 2002; Peterson 1981; Tiebout 1956). Finally, Orfield’s measure of the percentage of non-Asian minorities and percent foreign-born in 2010 were also used to determine the sociopolitical impacts of race or ethnicity on housing characteristics ⁴ .

Following Orfield’s (2002) analysis, a k-means cluster analysis was calculated for all zipcodes in the CMPS dataset. ⁵ In short, a cluster analysis is designed to group observations on the basis of similarity across the eleven aggregate-level measures including service environment (six), tax capacity (three), and race/ethnicity variables (two), resulting in a predefined set of homogeneous clusters or groups. These cluster definitions were then mapped in a geographic information system (GIS) to determine the appropriateness of the groupings in terms of their spatial distributions. By comparing across the nation both the spatial locations of and the average values for each of the seven groups, we were able to designate each zipcode as one of the following types: very affluent job center, affluent job center, bedroom developing, at-risk low density, at-risk segregated, at-risk older, and central city (Orfield 2002). ⁶ Table 1 presents these aggregate-level characteristics, summarized for each cluster. ⁷

OPEN IN VIEWER

Table 1.

Characteristics of the Geographic Types.

Eleven aggregate level variables used to create k-means clusters
	Populationgrowth (%)	Median HHincome ($)	Median HHrent ($)	Medianhome value($)	Unemployed(%)	Poverty(%)	Populationdensity	Housing unitdensity	Medianstructure age	Non-Asianminority (%)	Foreign born(%)
Geographic types
Very affluent	17	98,604	1,539	836,935	4	9	20,446	10,697	1961	20	28
Affluent	9	78,849	1,397	609,121	5	11	14,599	6,133	1965	30	30
Bedroom developing	29	64,994	1,153	344,820	6	12	8,711	3,305	1970	45	26
At-risk low density	22	50,887	881	166,585	6	16	2,887	1,211	1979	43	14
At-risk old	35	62,989	1,047	245,691	6	13	4,701	2,085	1976	36	18
At-risk segregated	20	36,868	704	96,262	7	24	2,704	1,172	1968	54	10
Central city	232	71,089	1,282	457,808	5	12	12,493	4,696	1966	43	30

Source. 2016 Collaborative Multiracial Post-Election Survey (CMPS)/2010 U.S. Census, five-year ACS.

HH = household; ACS = American Community Survey.

Central cities, defined here, are areas that generally lack the resources to meet the physical and social needs of the residents and business. For decades after WWII, most central cities witnessed a mass exodus of families who opted for single family homes in suburban areas. Recently, central cities have had a resurgence in desirability, especially by young professional types. Despite this resurgence in popularity and desirability, many central cities still lack the resources to meet needs of the community. Schools and basic social services lack funding from decades of lost tax revenues and revenue allocation to the periphery to support the rapidly growing suburbs.

Three types of at-risk communities were examined, including at-risk segregated, at-risk older, and at-risk low density. Overall, at-risk geographic types experience significant fiscal or social stress and are susceptible to rapid decline, yet lie outside of the central city. Similar to Orfield’s (2002) findings, the data in Table 1 also indicate that in at-risk communities, median household values (tax capacities) are, on average, lower than other geographic types. In these communities, there are high levels of social need but limited or declining resources at the local level. In other words, lower home values, for example, make it difficult for governing jurisdictions to extract the tax revenues needed to provide the high-quality local goods and services within the community. These communities often lack substantial business districts, recreational and cultural attractions, and public infrastructure, and “as a result, these communities often become poor faster and lose local business activity even more rapidly than the cities they surround” (Orfield 2002, 37). A declining local tax base combined with high social needs can push tax rates up and/or services down.

On average, at-risk segregated communities have low tax capacities, slow growth, and high expenditure costs. At-risk older suburbs contain housing that is also located in or near the inner-ring city but that tends to house larger white populations than at-risk segregated suburbs. At-risk low-density geographies have tax bases that grow at rates slower than average and may have relatively high rates of poverty. However, these areas are distinct from at-risk segregated and at-risk older locals because they are located in more exurban locations, farther away from central cities, and may be experiencing population growth.

Bedroom-developing geographies can be characterized as the prototypical suburb: density is low, there are newer housing developments, and the tax capacity is just below average and growing at an average rate. Although these communities have only modest fiscal resources at their disposal, they do not yet face the social and fiscal stresses experienced by at-risk communities. The tax revenues generated here are able to better support the needs of the residents.

Finally, affluent and very affluent job centers are also characterized as “edge cities,” defined as “suburban communities with more than five million square feet of office space and more jobs than bedrooms” (Orfield 2002, 44; also see Garreau 1991). These areas benefit from high tax bases and very low costs compared with central cities, with concentrations of office and commercial space. The residents in these communities tend to be the most highly educated and wealthiest of all the geographic types.

We collapse some of these various community types for analytic and presentation clarity. We group very affluent and affluent into one category. We also group at-risk segregated and at-risk older into one at-risk category. As noted, at-risk low-density areas are located in more exurban locations, farther away from central cities, and may be experiencing population growth. Thus, preserve at-risk low density, central cities, bedroom developing. In all, this gives us five unique geographic types with which we can use to understand how residing in these communities may influence attitudes toward undocumented immigrants. ⁸ Table 2 shows the percentages of the final clusters as well as the racial breakdown by cluster.

OPEN IN VIEWER

Table 2.

Race/Ethnicity by Cluster Group.

Race	Cluster	% in Cluster
White	Central city	5.88
White	Affluent	8.33
White	At risk	43.24
White	Bedroom developing	14.61
White	Low density	27.94
Latino	Central city	8.71
Latino	Affluent	7.19
Latino	At risk	41.44
Latino	Bedroom developing	16.40
Latino	Low density	26.26
Black	Central city	5.82
Black	Affluent	3.22
Black	At risk	51.58
Black	Bedroom developing	9.73
Black	Low density	29.65
AAPI	Central city	16.56
AAPI	Affluent	23.76
AAPI	At risk	25.28
AAPI	Bedroom developing	18.72
AAPI	Low density	15.68

Source. 2016 Collaborative Multiracial Post-Election Survey (CMPS)/2010 U.S. Census, five-year ACS.

AAPI = Asian Americans and Pacific Islander; ACS = American Community Survey.

Findings and Discussion

First, we consider a model that incorporates our measures of geographic context, economic views, political predispositions, racial/ethnicity identity, and other socio-demographics. One of the advantages of the CMPS is the large samples of blacks, Latinos, AAPIs, and whites. While the focus of this study is on the role of geographic identity, we also seek to understand whether these results are consistent when considering other identities such as a respondent’s race or ethnicity. In Table 3, our full model examines individual-level attitudes toward undocumented immigrants, using ordinary least squares (OLS) regression. ¹⁰ In Table 3, we examine our results for models separated by geographic types.

OPEN IN VIEWER

Table 3.

Views toward Undocumented Immigrants Currently Living and Working in the United States.

	Views toward undocumented immigrants
Geographic type
Affluent	−0.417** (0.137)
At risk	−0.612*** (0.116)
Bedroom seveloping	−0.490*** (0.126)
Low density	−0.854*** (0.118)
Economic views
Economic assessment	−0.128** (0.049)
Context
Latino growth	0.014* (0.006)
Political predispositions
Democrat	0.739*** (0.069)
Independent	0.281*** (0.059)
Ideology	−0.333*** (0.025)
Racial identity
Black	0.652*** (0.090)
Latino	1.064*** (0.092)
AAPI	−0.269* (0.117)
Socio-demographics
Education	0.034 (0.023)
Income	0.044*** (0.009)
Born again	0.203*** (0.058)
Female	0.274*** (0.049)
U.S. born	−0.011 (0.086)
Married	−0.142** (0.054)
Age	−0.003 (0.001)
Observations	8,223
Log likelihood	−6,856.366

Source. 2016 Collaborative Multiracial Post-Election Survey (CMPS).

AAPI = Asian Americans and Pacific Islander.

An asterisk indicates that the variable is statistically significant: *p < .05. **p < .01. ***p < .001. Values in parentheses refer to standard errors.

In Table 3, we set central city as the base category in order to interpret the association of our new geographic measures relative to respondents residing in a central city. We see this as appropriate since much of the existing literature has considered urban versus rural contexts, and we seek to highlight our findings relative to the existing literature.

Examining each of the four geographic types, we see that each of these negatively associate with attitudes toward undocumented immigrants. Residents in affluent geographic types are significantly less likely to have favorable attitudes toward undocumented immigrants compared with their counterparts in the central city. This statistically significant relationship holds for those in at-risk, bedroom developing, and low-density areas, even controlling for other existing explanations. Table 3 provides positive support for our hypothesis that one’s geography is predictive of attitudes toward undocumented immigrants.

To get a better idea of the substantive effects and how these geographic communities impact these immigration attitudes vis-à-vis one another, we use predicted probabilities. Figure 2 shows the predicted probability of support for each possible value of the dependent variable. In the left panel, we predicted the probability that an average respondent living in each of the different community types would say that undocumented immigrants in the United States should leave immediately. Along the x-axis is each of the community types. The results in the first panel show three important takeaways. First, those living in the central city are the least likely to think that undocumented immigrants should leave immediately. In contrast, those in the low density are the most likely to support this policy stance. In fact, those living in low-density areas are almost twice as likely to think that undocumented immigrants should leave immediately compared with those living in the central city. Those living in the areas affluent, at risk (older and segregated), and bedroom developing suburbs hold positions in between the central city and low-density residents.

OPEN IN VIEWER

Figure 2.

Predicted attitudes toward undocumented immigrants currently living and working in the United States, by geographic type.

In the center panel, we show the probability of responding that undocumented immigrants in the United States should stay, but only temporarily. Here, we see a consistent pattern with the first panel. Those living in the central city are the least likely to support this policy position, whereas those living in the low-density areas are the most supportive. Finally, in the third panel, we show the probability of responding that undocumented immigrants in the United States should be allowed to stay and allowed to apply for U.S. citizenship. Unlike the other two policy positions, here we see the opposite. Those living in the central city are the most supportive of this policy. Those living in low-density areas, however, are the least supportive. Consistent with the other policy positions, those living in affluent, at-risk, and bedroom developing communities lies somewhere in the middle. Respondents in each of these community types are considerably less supportive of this policy position than those in the central city but considerably more supportive than similarly situated counterparts in low-density areas.

Examining our controls for Latinos, AAPIs, and blacks (making whites the reference category), blacks and Latinos are significantly more likely to have favorable attitudes toward undocumented immigrants compared with their white counterparts. AAPIs, however, are significantly less likely to have favorable attitudes toward undocumented immigrants. These findings are aligned which recent data regarding black and Latino support for undocumented immigration. Following the announcement of Deferred Action for Parents of Americans and Lawful Permanent Residents (DAPA), Gallup polling found that a slight majority (51%) disproved of Obama’s executive actions on immigration. However, both non-Hispanic blacks and Latinos supported the actions 2 to 1. In fact, a slightly higher percentage of non-Hispanic blacks supported Obama’s executive actions on immigration at 68 percent, followed by Latinos at 64 percent and non-Hispanic whites at 30 percent. Of all immigrants, 69 percent supported Obama’s executive actions on immigration (Jones 2014).

Next, we subset each of the analyses by geographic type. We seek to understand how the traditional predictors of attitudes toward immigrants vary by place, beyond an urban/rural or urban/suburban dichotomy. We present these results in Table 4. For each column in Table 4, we subset the data based on community type and regress attitudes toward undocumented immigrants on the control variables used in the other models, which represent a set of traditional predictors for political attitudes. Columns are ordered by central city, affluent, at risk, bedroom developing, and low density. Beginning with assessments of the economy, which is coded where higher values are respondent’s feeling the economy is getting worse, we see some stark patterns given the type of community. For those who live in central cities, there is no relationship between thinking the economy is getting worse and attitudes for immigrants. In fact, there is no relationship in at-risk and bedroom developing communities. We only find significant relationships in affluent and low-density community types. In both areas, we find that those who think the economy is getting worse are significantly less likely to hold favorable attitudes toward undocumented immigrants.

OPEN IN VIEWER

Table 4.

Views toward Undocumented Immigrants Currently Living and Working in the United States, by Geographic Type.

	Views toward undocumented immigrants
	Central city	Affluent	At risk	Bedroom developing	Low density
Economic views
Economic assessment	−0.020 (0.263)	−0.471* (0.210)	0.018 (0.075)	−0.047 (0.138)	−0.355*** (0.091)
Context
Latino growth	−0.024 (0.031)	−0.099** (0.034)	0.018* (0.009)	0.019 (0.017)	0.010 (0.010)
Political predispositions
Democrat	−0.585 (0.360)	0.700** (0.245)	0.934*** (0.106)	0.396* (0.194)	0.814*** (0.131)
Independent	−1.397*** (0.368)	0.027 (0.241)	0.438*** (0.093)	0.562** (0.174)	0.198 (0.103)
Ideology	−0.636*** (0.140)	−0.425*** (0.095)	−0.344*** (0.039)	−0.485*** (0.068)	−0.216*** (0.047)
Racial identity
Black	0.428 (0.446)	0.257 (0.463)	0.552*** (0.128)	0.611* (0.296)	0.895*** (0.163)
Latino	1.072* (0.423)	1.466*** (0.418)	1.114*** (0.139)	0.075 (0.222)	1.600*** (0.191)
AAPI	−0.606 (0.383)	−0.746* (0.361)	0.072 (0.208)	−0.811** (0.281)	0.032 (0.261)
Socio-demographics
Education	−0.0002 (0.117)	0.357*** (0.085)	−0.005 (0.034)	−0.147* (0.062)	0.088* (0.045)
Income	−0.195*** (0.045)	0.086* (0.035)	0.047*** (0.014)	0.048* (0.022)	0.073*** (0.017)
Born again	2.029*** (0.476)	−0.636* (0.274)	0.468*** (0.091)	−0.288 (0.169)	0.095 (0.100)
Female	1.442*** (0.275)	0.155 (0.199)	0.139 (0.076)	0.332* (0.138)	0.361*** (0.093)
U.S. born	0.16 (0.331)	−0.370 (0.362)	0.235 (0.133)	−0.581** (0.205)	−0.049 (0.180)
Married	0.512 (0.310)	−0.882*** (0.231)	−0.074 (0.084)	0.234 (0.147)	−0.332*** (0.097)
Age	−0.044*** (0.009)	−0.015** (0.006)	−0.004 (0.002)	0.007 (0.004)	0.005 (0.003)
Observations	545	681	3,441	1,232	2,324
Log likelihood	−290.291	−509.053	−2,872.875	−965.038	−2,051.692

Source. 2016 Collaborative Multiracial Post-Election Survey (CMPS).

AAPI = Asian Americans and Pacific Islander.

An asterisk indicates that the variable is statistically significant: *p < .05. **p < .01. ***p < .001.

Next, we consider the role of partisanship. In the full model, partisanship was positively related to attitudes toward immigrants. However, examining the patterns by community types, we see that pattern holds in affluent, at-risk, bedroom developing, and low-density areas. The relationship does not hold in Central Cities. While it is not statistically significant, the direction of the coefficient is negative. Looking at the rest of the results, we see that these factors operate differently given the community type, providing more evidence that the role of one’s community type is important in understanding important political attitudes.

In terms of racial and ethnic identity, Table 4 shows that attitudes toward immigrants vary considerably across racial and ethnic groups by community type, providing further evidence of the role of place in important political attitudes. Since the base category is whites, the coefficients shown in Table 3 are compared with whites. Taking the first row, across the board, black’s attitudes toward immigrants are always positive, but only statistically significant in at-risk, bedroom developing, and low-density communities. In the central city and affluent areas, these effects are positive, but not statistically significant. This suggest that in these areas, we can fail to reject that alternative hypothesis that black have attitudes distinguishable from their white counterparts even when controlling for a number of standard individual controls.

Latinos and AAPIs show distinct patterns of attitudes toward undocumented immigrants. Latinos, except in bedroom developing communities, always show positive and statistically significant relationships. This means that compared with whites, Latinos in these areas are much more supportive of immigrant than their white counterparts. In bedroom developing communities, the relationship is positive, but not statistically significant.

Among AAPIs, the results are the most mixed. In affluent and bedroom developing communities, there is a statistically significant and negative relationship, suggesting that compared with whites, AAPIs in these communities show less favorability toward undocumented immigrants. In two other community types (at risk and low density), the relationship is positive, but not near statistical significance. In the central city, AAPIs have lower but not statistically distinguishable attitudes toward undocumented immigrants.

We also consider two additional sets of models where we interact racial/ethnic group membership with assessments of the economy and a set of models where we interact Latino growth with racial/ethnic group membership. We present the full models in the Online Supplemental Appendix, but briefly comment on some key findings here. More negative assessments of the economy among whites (base term) is associated with lower support for immigrants, but only in affluent areas and low-density areas. Compared with whites, blacks and Latinos with high evaluations of the economy (black and Latino base term) are more supportive of immigrants across the neighborhood type. For blacks, the relationship is positive and statistically significant in at-risk, bedroom developing, and low-density areas. For Latinos with positive economic assessments, there is a positive and significant relationship across the community types except for bedroom developing. Supporting the results from Table 3, AAPIs with positive views of the economy still harbor unfavorable attitudes toward undocumented immigrants. Regarding the interaction terms, which show the relationship among racial/ethnic group members with low evaluations of the economy, only Latinos in the central city show a statistically significant relationship. This means that Latinos with poor assessments of the economy in central city communities harbor more unfavorable attitudes toward undocumented immigrants than their white counterparts.

We also examine how Latino Growth interacts with racial/ethnic membership across various community types in Online Supplemental Appendix Table 2. The main takeaways include negative and statistically significant views of undocumented immigrants among whites as the Latino populations grows, but only in affluent and central city areas. In at-risk areas, there is a substantively small but positive relationship. Among Latinos, the results are mixed between those who live in areas with little growth (base term) and those who reside in areas with growth (interaction term) across the different community types. While results are always positive, conditional on Latino growth, the overall ability to statistically distinguish the effect from whites under low growth is not possible. What this means is that community type has a strong impact on attitudes across racial and ethnic groups. AAPIs show little favorability toward immigrants in low-growth central city and affluent areas, but significant and positive support in high-growth central city and affluent areas, suggesting that contact and increased community connections of a shared immigrant experience may help forge more positive attitudes. Our analysis also shows some heterogeneity among Latinos. For example, while there is consistent support for immigrants among Latinos in general, those living in bedroom developing communities harbor attitudes undocumented immigrants that are closer to whites.

Discussion and Conclusion

This analysis moves beyond the metro/non-metro dichotomy, providing a more nuanced understanding of the types of communities that people reside and how residing in those communities influence political attitudes. Attitudes toward undocumented immigrants are not uniform across various kinds of geographic environments, suggesting that these contexts impact how people interpret and understand contemporary political phenomena. Contexts and environments are powerful forces that shape many of social and political attitudes (Collingwood et al. 2018; Perez 2015; Reny, Wilcox-Archuleta, and Nichols 2019; Vargas et al. 2017).

Recent work has looked at various social categories and categorizations across various dimensions including geography as politically important group-based identities (Achen and Bartels 2016; Garcia-Rios, Pedraza, and Wilcox-Archuleta 2018; Hochschild 2016; Cramer 2016; Mason 2015). Other scholars have noted how experiences with different types of discrimination are linked to divergent political outcomes (Oskooii 2016, 2018), which are likely mediated by context and geography. We see the idea of a shared geographic-based identity, not solely focused on rural residents as Cramer’s (2016) recent work suggests, as important for future research. Our results suggest that attitudes toward undocumented immigrants vary considerably across these geographic types. Future work should consider mechanisms by which these geographic types impact political attitudes toward other racialized and non-racialized policies.

We also contribute to the growing body of work that considers how space is defined. In many cases, scholars have focused on how one or two key characteristics of context are related to an outcome of interest. Our approach considered a number of factors to classify a type of community. We think this approach will be of interest to other scholars who want to consider how spatial arrangements are systematic and follow detectable patterns of similarities and difference. The use of the k-means clustering allowed us to consider a full set of multi-dimensional local characteristics in a highly reduced single dimension. Our typology of place-based identity is one step toward a greater understanding of how one’s identity can influence attitudes toward undocumented immigrants as well as shape public opinion toward a host of other racialized and non-racialized policies at the federal, state, and local level.

Supplemental Material