The 2000s have witnessed an expansion of interior immigration enforcement in the United States. At the same time, the country has experienced a major demographic transformation, with the number of U.S. citizens living in mixed-status households—that is, households where at least one family member is an unauthorized migrant—reaching 16 million. U.S. citizens living in mixed-status households are personally connected to the struggles experienced by their unauthorized family members. For them, immigration policy is likely to shape their current and future voting behavior. Using data from the 2002–2014 Current Population Survey Voting and Registration Supplements, we examine whether intensified immigration enforcement has affected the political engagement of U.S. citizens living in mixed-status households. We find that immigration enforcement has chilled their electoral participation by lowering their propensity to register by 5 %; however, it has not visibly affected their voting propensity among those registered. Importantly, their lower voting registration likelihood does not seem to reflect indifference for community and public matters, given that it has been accompanied by greater involvement in civic forms of political participation, such as volunteering. Understanding how immigration policy affects the political participation of a fast-growing segment of the electorate is imperative because they will inevitably constitute a rapidly rising political force in future elections.
The 2000s have witnessed an impressive expansion of interior immigration enforcement. Between 2003 and 2013, funding for the Immigration and Customs Enforcement (ICE) agency—the U.S. federal agency responsible for interior immigration enforcement—increased approximately 80 %, apprehensions more than doubled, and removals increased threefold.1 The increase in the intensity of interior immigration enforcement is also evidenced in the greater participation by state and local governments in a number of immigration enforcement initiatives and programs.
At the same time, the United States has found itself in the midst of a major demographic transformation. Since 2000, the number of U.S.-born children with at least one unauthorized parent has more than doubled, with more than 4.5 million U.S. citizen children living with unauthorized parents (Passel and Cohn 2011; Taylor et al. 2011). It is also estimated that more than one-half million unauthorized migrants have a U.S. citizen spouse (Pastor et al. 2014). Although U.S. citizens living in mixed-status households are not the targets of immigration enforcement,2 they are personally connected to the struggles that their unauthorized family members experience. These struggles include increased stress and anxiety over familial separation, geographic relocation in order to evade the apprehension of a family member, or a significant loss of household income when a family member (typically the household head) is deported. Thus, for the more than 16 million individuals living in mixed-status families, immigration enforcement is a key policy issue that is likely to shape their current and future voting behavior.
In this study, we examine the effect of intensified enforcement on the political engagement of U.S. citizens living with unauthorized family members. Using data from the 2002–2014 Current Population Survey (CPS) Voting and Registration Supplements, a proxy for immigrants’ unauthorized status to distinguish citizens in mixed-status households, and detailed information on the timing and geographic scope of interior enforcement policies, we examine how intensified immigration enforcement is affecting the registration and voting behavior of U.S.-born citizens living with a likely unauthorized household member. On one hand, their political engagement might increase if they have a strong desire to express their discontent with residing in a household subjected to the fear, anxiety, and financial hardships associated with an increase in enforcement. On the other hand, the hardships may cause them to become politically disengaged and disenfranchised and lower their political participation. Either way, U.S. citizens living with unauthorized immigrants have a personal connection to immigration policy and to the struggles that unauthorized migrants endure, which can potentially affect their political participation (Ewing and Cantor 2014).
Using a constructed population weighted index of the intensity of interior immigration enforcement at the metropolitan statistical area (MSA) level, we show that the intensification of interior immigration enforcement chills the political participation of U.S. citizens in mixed-status households by lowering their likelihood of registering to vote by 3.6 percentage points (5.3 %). However, the voting likelihood of those registered is not significantly altered. The negative effect of intensified immigration enforcement on the registration likelihood of citizens in mixed-status households seems to operate through police-based measures that are directly linked to the apprehension and deportation of their loved ones, suggesting that these U.S. citizens might be reticent to register with government authorities. Furthermore, the effect seems to be concentrated among U.S. eligible voters in mixed-status households who are men, older than age 23, and of non-Mexican descent. These findings corroborate the known greater voting propensity of women, the extensive voting registration drives on college campuses, and the well-established roots and greater mobilization of Mexican migrants in the United States (e.g., Barreto and Woods 2005; Lopez and Marcelo 2008; White 2016).
We also look at other forms of political engagement, such as volunteering in general as well as in specific groups, including political and advocacy groups. Our intent is to explore whether this demographic, perhaps afraid of registering with a government authority, has turned to other forms of community and public involvement that they might perceive as less risky (in the sense that these forms do not require them to register with a government authority) and that simultaneously present positive externalities. The community might serve as some sort of safety net for their loved ones in the face of intensified enforcement. Although we are constrained by data availability, we find that these same individuals seem to have also become significantly more likely to engage in their communities by increasing their volunteering activities, including those involving community, immigrant, political, and advocacy groups, anywhere between 17 % and 24 %. Thus, it does not appear that they have become politically disengaged; rather, they have become disengaged from the formal electoral process.
Overall, the findings uncover some of the hidden costs of intensified enforcement on U.S.-born citizens living in mixed-status families. This is an important area of research given that an estimated 2.25 million of the 6.3 million U.S. citizens living in the same household as a relative who would have been eligible for Deferred Action for Parents of Americans and Lawful Permanent Residents (DAPA) will be eligible to vote by 2020 (Pastor et al. 2015).3 Although political participation can take various forms, such as participating in registration drives or voting rallies, the most important expression in a democratic society is through registering to vote and voting. Ensuring that such participation takes place is vital to the health of a democracy. Because U.S. citizen children with unauthorized parents represent one of the fastest-growing segments of the electorate, as well as one exposed to the hardships of unauthorized immigrants, this demographic inevitably constitutes a rapidly rising political force with the potential to sway future national, state, and local elections.
Immigration Enforcement and Voting Behavior
In response to the intensification of enforcement, a growing body of work has examined the effect of immigration policies on unauthorized immigrants and/or their families.4 Watson (2014) documented that heightened federal immigration enforcement leads to “chilling effects” in Medicaid participation among children of noncitizens, even when the children are U.S. citizens. Amuedo-Dorantes and Lopez (2015) found evidence to suggest that the greater immigration enforcement has a negative effect on the schooling progression of children with unauthorized parents. The increase in deportations accompanying the intensification of interior immigration enforcement has also been shown to break up families and cause children to be left behind in the care of a single parent, an older sibling, or other relatives (Amuedo-Dorantes and Arenas-Arroyo 2017; Landale et al. 2011). Even when immigration enforcement measures do not specifically target employment, they can induce families to start living in the shadows because of fear of greater exposure to detection and apprehension—a decision that negatively affects employment and earnings opportunities. We contribute to this growing body of work by examining the effect of immigration enforcement on political behavior.
Because they personally connect to the struggles that their parents, spouses, or other unauthorized family members face, eligible voters in mixed-status households might have a strong desire to express their discontent with how society is reacting to and treating immigrants. Between 1996 and 2013, approximately 5 million noncitizens were deported (Simanski 2014). Because many of those immigrants had families in the United States, including spouses and children who are U.S. citizens, deportations resulted in family separations with grave consequences (Ewing and Cantor 2014). As a result, it is not surprising that Lopez and Marcelo (2008) found that U.S.-born youth with foreign-born parents are twice as likely as U.S.-born youth with U.S.-born parents to participate in protest activities to express their discontent with living in households afflicted by fear, anxiety, and financial hardships associated with the intensification of immigration enforcement. Bloemraad and Trost (2008) also found that a significant amount of youth participated in the surge in immigrant rights protests that occurred during the spring of 2006: these U.S.-born youth had more direct access to mainstream media and institutions and political information than did their immigrant parents. Likewise, Barreto and Woods (2005) examined voter turnout among Latinos residing in Los Angeles County between 1994 and 1998. They found that Latinos responded to anti-immigrant and anti-Latino legislation, such as Propositions 187, 209, and 227, by increasing voter turnout. Barreto et al. (2005) also examined the effect of the 1986 Immigration Reform and Control Act (IRCA) on Latino voter turnout in California. They found that IRCA did not cause the increase in Latino voting in the 1990s; rather, demographic changes and mobilization efforts led to such an increase. Similarly, also hinting on the importance of mobilization efforts, White (2016) found an increase in county-level Latino turnout rates in the wake of the adoption of the Secure Communities program.5 White (2016) compared the 2006–2010 change in aggregate voter turnout rates in reluctantly enrolled jurisdictions (namely, late adopters of the program in five states that had universally adopted the program) to the change in non-enrolled jurisdictions. To the extent that the findings refer to the wake of the program, the result is interpreted as more evidence of mobilization in response to threat. Finally, Holbrook et al. (2016) found that Latinos are more likely to engage in activism to express opinions about immigration compared with whites and blacks.
Yet, the aforementioned hardships could plausibly cause eligible voters to distrust government, becoming politically disenfranchised (Krogstad and Lopez 2014). Rocha et al. (2015) used a public opinion survey of Texas residents to examine the effect that deportations through Secure Communities have on trust in government (both local and federal) and political efficacy among Latinos and non-Hispanic whites. They found that both native and foreign-born Latinos experience negative political orientations in areas where deportations are higher. In 2008, less than one-half of the Hispanics surveyed in a Pew Research Center study had confidence in the police and courts (Lopez and Livingston 2009). In fact, this pattern supports the finding of lower participation among Hispanics in past elections (Pew Research Center 2012).
The effects of enforcement on political participation might also be highly heterogeneous. DeSipio (2013) found that naturalized Latinos are 29 % less likely to vote than U.S.-born Latinos. However, among the naturalized, he found that individuals who report naturalizing for political reasons are 9 % more likely to vote than other naturalized Latinos. If immigrants choose to become U.S. citizens for political reasons, such as to participate in the political process, their political engagement may not be as negatively affected by enforcement relative to other eligible voters.
Women are also typically more likely to register and vote than men (e.g., Schlozman et al. 1995), and they have different positions on key election issues than men (Whitaker 2008). Thus, women’s political engagement may be less adversely affected by enforcement relative to male eligible voters.
Hispanics are less likely to register and vote than non-Hispanic whites (Pew Research Center 2006). At the same time, Hispanics—particularly Mexicans—have more established political roots in the United States due to historic migration patterns. In addition, Mexican immigrants, both naturalized and noncitizen, are just as likely to engage in nontraditional political activities. This engagement could set a precedent for increasing their likelihood of voting when eligible or affect more formal acts of political participation, such as voting and registration, among eligible voters living in mixed-status households (Barreto and Muñoz 2003; Tran 2016).
Finally, youth are typically the least likely to vote—a finding that also holds true for Latino youth (Lopez and Gonzalez-Barrera 2013). Latino youth may be less inclined to register or vote given their parents’ lack of familiarity with the U.S. political system or their lesser political involvement owing to deportation fears associated with intensified enforcement or their disenchantment with politics (Humphries et al. 2013). Yet, the number of youth participating in political protests is also significant (Bloemraad and Trost 2008). For Latinos, the greater participation may be due to strong connections to their immigrant families; ethnic identities; or connections to issues affecting their immigrant peers, such as the in-state tuition movement (Sief 2011). In addition, the majority of the children of immigrants are U.S.-born and educated in U.S. schools, which some have argued can directly or indirectly incentivize their civic engagement to a larger degree than among their parents (Barreto and Muñoz 2003; Humphries et al. 2013). Finally, registration drives on college campuses facilitate the participation of younger eligible voters when compared with their older eligible voters.6 As a result, enforcement may have less of a negative effect on young eligible voters.
To our knowledge, this is the first study to systematically assess the effects of intensified immigration enforcement on traditional forms of political participation, such as voting and registration, and less traditional types of political engagement, such as volunteerism. This nuance is important given research showing that non-electoral political participation is a viable option for many ethnic groups and one that can help enhance electoral political participation at a later point (Barreto and Muñoz 2003; Tran 2016).
Data Sources and Descriptive Statistics
Data on Political Participation
We rely mainly on two data sources: (1) the 2002, 2004, 2006, 2008, 2010, 2012, and 2014 CPS November Voting and Registration Supplements; and (2) local- and state-level data on the enactment and implementation dates of numerous measures of interior immigration enforcement adopted since 2002.
The November Voting and Registration Supplements ask respondents whether they registered and voted in the most recent November election. The survey has been conducted for every congressional and presidential election year since 1964. We use the responses to those questions to capture their political participation. Because the CPS data lack sensitive information on the immigration status of migrants, we rely on ethnicity, citizenship, and educational attainment—traits shown to be good predictors of immigrants’ unauthorized status (Passel and Cohn 2009, 2010)—to proxy for the likely unauthorized status of household members. In addition, because most nonimmigrant visas for low-skilled workers are for short periods that do not exceed five years including renewals, we further restrict the definition of likely unauthorized to Hispanic noncitizen household members who have, at most, a high school education and have resided in the United States in excess of five years.7 Using all these traits, we obtain an estimated unauthorized immigrant population of 12,791,033 immigrants—a figure that is very close to the estimated population of 11 to 12 million undocumented immigrants in the United States using the residual method.
The fact that the combination of these descriptors does a good job when trying to proxy for the likely undocumented status of immigrants is not unexpected. First, the U.S. Census Bureau and the U.S. Department of Homeland Security estimated that in 2012 nearly 40 % of noncitizens are authorized immigrants (Acosta et al. 2014; Baker and Rytin 2013). That is, noncitizens include all unauthorized immigrants as well as many authorized immigrants. Second, because of geographic proximity and poor economic and social conditions at home, as well as extensive migrant networks, more than two-thirds of unauthorized immigrants in the United States are Hispanic from Mexico and Central America. Hence, we follow the convention in the literature of adopting Hispanic noncitizens as a proxy for individuals who are likely to be undocumented (Bohn and Pugatch 2013; Orrenius and Zavodny 2016; Passel and Cohn 2009; Pope 2016). Third, as noted in previous research (see, for example, Bohn and Lofstrom 2013; Orrenius and Zavodny 2016), most unauthorized immigrants have relatively little education because they are from countries with low average levels of educational attainment. Approximately three-quarters of adult unauthorized immigrants have no more education than a high school diploma (Passel and Cohn 2009). Finally, to address any concerns regarding the possibility that our sample might include low-skilled immigrants or college students with nonimmigrant visas, we also restrict our analysis to Hispanic noncitizen individuals who have not completed high school and have lived in the United States five or more years. This last restriction further ensures that the low-skilled migrant is not legally in the United on a nonimmigrant visa, which is typically granted for a much shorter duration.8 Given that our proxy is attempting to measure the likely immigration status of an individual, we adopt the term “likely unauthorized.” As with any proxy, it may include false positives and false negatives.
Using the imputed likely undocumented status of immigrants in the sample, we create a dummy variable indicative of living in a mixed-status household. The latter is defined as households with at least one likely unauthorized immigrant, whether that member is a parent, a spouse, or another family member, such as an older sibling who migrated to the United States at an early age. Nevertheless, the vast majority of the unauthorized immigrant household members in our sample are either the parents or spouses of U.S.-born citizens.9
Enforcement Data and Its Temporal and Geographic Variation
We gather information regarding the timing and geographic scope of various interior enforcement policies. Specifically, data on the enactment of state-level employment verification (E-Verify) mandates—often a key element in the omnibus immigration laws (OIL)—and data on OIL are gathered from the National Conference of State Legislatures (NCSL) website.10 Data on the implementation of 287(g) agreements and Secure Communities (SC) at the state and local levels are collected from the ICE 287(g) Fact Sheet website, from Kostandini et al. (2014), and from the ICE’s Activated Jurisdictions document, respectively (U.S. Immigration and Customs Enforcement 2012a, b, c). We use these data to construct a weighted MSA-level index that captures the intensity of immigration enforcement.11 Details on the construction of this index can be found in section B of Online Resource 1.
There are several advantages to using a single index of enforcement as a proxy for the intensification of immigration enforcement.12 First, an index provides a more comprehensive way of measuring the overall effect of a variety of immigration enforcement initiatives and, to the extent that any one measure can be applied differently by any two entities, distinguishing across the various types of measures does not necessarily shed much light. Instead, what we do in later heterogeneity analyses is distinguish between police-based and employer-based enforcement measures given their distinct consequences, the former being more directly linked to apprehension and deportation.13 Second, through the combination of various indexes—each weighted by the population and number of months the measure was in place in that particular year—the index allows us to capture the depth and intensity of immigration enforcement in a given area, as opposed to only whether enforcement existed. Third, immigration enforcement is an interconnected system administered by various federal, state, and local authorities and agencies with similar missions, and some measures, such as Secure Communities, were enacted as a continuum of prior existing measures, like the 287(g) program. The index allows us to better address this interconnectedness while facilitating the interpretation of the overall impact of intensified enforcement.14
Figure 1 illustrates the progressive rollout of immigration enforcement measures throughout the country over the period under consideration. For simplicity, we collapse the MSA enforcement index at the state level and display the level of interior immigration enforcement at three 5-year intervals starting two years after the adoption of the first enforcement measures: 2004, 2009, and 2014. Atop an increase in interior enforcement between 2004 and 2009, the maps in Fig. 1 show an accelerated increase in the 2009–2014 period.
Tables 1 and 2 present some summary statistics for our sample of eligible voters for the 2002–2014 period. Approximately 1 % of U.S. citizens aged 18 or older reside in a mixed-status household (Table 1).
There are striking differences in the share of eligible voters who register and vote in households without and with a likely unauthorized immigrant. On average, 67 % of eligible voters in households without a likely unauthorized immigrant indicate having registered in the most recent November election, relative to 40 % in mixed-status households (Table 2). Of those registered to vote, 77 % of those residing in households without a likely unauthorized immigrant indicate having voted in the last November election, as opposed to 64 % in mixed-status households (Table 2). As shown by the t tests in the last column of Table 2, these differences are statistically significant.
In addition, Tables 1 and 2 uncover some important demographic traits of our samples of eligible voters. On average, eligible voters in mixed-status households are significantly younger than their counterparts in households without a likely unauthorized adult (35 years old vs. 46 years old, respectively). Approximately 50 % are female in both samples. However, the vast majority of eligible voters in mixed-status households are Hispanic (94 %); among those in households without a likely unauthorized immigrant, that share is about 13 %.15 Linked to their younger age, a significantly larger share of eligible voters in mixed-status households are single than in households without a likely unauthorized immigrant (56 % vs. 48 %, respectively). The educational attainment of those in mixed-status households is also significantly lower than that of their counterparts in non-mixed-status households. For instance, on average, only 29 % of eligible voters in mixed-status households have more than a high school education, relative to 63 % of their equals in households without a likely unauthorized immigrant.
Finally, the areas in which eligible voters reside also differ in some respects depending on whether respondents reside in a mixed-status household. The most noticeable one refers to the share of likely unauthorized in the MSA where they reside, which averages 10 % in the case of eligible voters in mixed-status households (vs. 7 % in the case of eligible voters residing in other households). Interestingly, a larger share of eligible voters in mixed-status households reside in a state with a Republican governor than their counterparts in households without a likely unauthorized adult (61 % vs. 57 %, respectively). Yet, as we might expect, eligible voters from mixed-status households reside in areas with a somewhat lower intensity of interior immigration enforcement than other eligible voters from households without a likely unauthorized immigrant. Similarly, a lower share of eligible U.S. voters in mixed-status households reside in states with voter ID laws when compared with eligible voters from other households. We account for these significant differences in our analysis.
Equation (1) includes the vector Xi,h,m,t, which accounts for individual-level characteristics known to be potentially correlated with the outcomes being examined, including age, gender, ethnicity, marital status, and educational attainment (e.g., Schlozman et al. 1995; Verba and Nie 1972; Verba et al. 1995). Additionally, Eq. (1) incorporates a number of MSA-specific and time-varying characteristics (included in Zm,t) that potentially influence our outcomes, including average household income by MSA—a proxy for wealth (Levin-Waldman 2013)—as well as the share of likely unauthorized migrants in an MSA in a given year. Additionally, we include information on the state housing price index (another measure of wealth), state-level unemployment rates for a given year, a dummy variable indicating whether the state in question had a Republican governor that year, and a dummy variable indicative of whether the state had specific ID requirements for voting. We also include an indicator for whether the past elections were presidential elections.
Equation (1) also includes MSA (ϑm) and year (μt) fixed effects as well as MSA-specific time trends (ϑmt). MSA and year fixed effects address unobserved and time-invariant local area and temporal characteristics correlated with political participation, such as residing in an economically depressed MSA or being surveyed during the 2008–2009 Great Recession. Finally, MSA-specific time trends capture a variety of unobserved time-varying characteristics at the MSA level not addressed by the controls included in Zm,t. Standard errors are clustered at the MSA level.
We face a number of econometric challenges in our analysis. One is the assumption of parallel trends in the political participation of youth: otherwise, any effects attributed to the intensification of immigration enforcement could prove to be preexisting and driven by something else. To assess this assumption, we use policy leads and placebo tests to explore for any anticipation effects or preexisting diverging trends.
A second challenge is the selection of respondents and areas into different levels of enforcement—potential sources of endogeneity via unobserved heterogeneity and reverse causality, respectively. Indeed, the fact that the average intensity of interior immigration enforcement is lower in the MSAs where eligible voters in mixed-status households reside is suggestive of the nonrandom location of respondents and/or the nonrandom adoption of tougher immigration enforcement by MSAs. To address these two concerns, we first assess what the response of eligible voters in mixed-status households would have been had they resided in MSAs where the countrymen of the likely unauthorized migrants in their household used to reside in 1980: that is, well before the implementation of any of the interior immigration enforcement measures under consideration. If mixed-status households strategically choose locations with a lower intensity of immigration enforcement, our estimates are lower-bound estimates of the effect that enforcement would have on the political participation of eligible voters in these households. However, if mixed-status households, for the most part, do not make their residential location decisions on the basis of the intensity of interior immigration enforcement—but perhaps on other factors, such as the existence of established networks of family and friends—our estimates might not be that different.
Subsequently, we address the nonrandom adoption of immigration enforcement by MSAs. Note that such adoptions do not need to be random. Yet, they should not be driven by area-specific characteristics correlated to the outcomes of interest. We test whether that is the case by examining whether political participation of those eligible to vote in mixed-status households in the area prior to the adoption of the enforcement measures was a significant determinant in the timing of adoption of the policies in question.
We also perform a number of robustness checks to check the sensitivity of our findings to the use of alternative samples, such as excluding Maricopa County in Arizona (an outlier MSA in terms of immigration enforcement) or excluding states that have passed a TRUST Act, thereby making the implementation of any immigration enforcement lax to null. Next, we proceed to explore the channels through which the effect of intensified immigration enforcement is likely to be taking place. The latter might include the fearful environment in which the household has lived following the intensification of police-based enforcement, which has been linked to deportations, and/or the restrictive employment and earnings opportunities available to undocumented immigrants and adversely affecting the households in which they reside. To conclude, we look at various demographic groups to learn about heterogeneous impacts by ethnicity, Mexican origin, age, and gender, which helps us gain a better understanding of the response to intensified immigration enforcement.
How Does Intensified Immigration Enforcement Affect Political Engagement?
Table 3 displays the estimated effect of increased immigration enforcement on the registration and voting of eligible voters living with a likely unauthorized parent or spouse. We estimate various model specifications that progressively add a number of individual-, MSA-, and state-level controls, as well as MSA/year fixed effects and MSA-specific time trends. Our first specification includes only the key regressors, along with MSA and year fixed effects, to assess the estimated impact of intensified enforcement within MSA and year. Subsequently, in specification (2), we add basic demographic characteristics of the eligible voter, including age, gender, ethnicity, marital status, and educational attainment. Specification (3) then incorporates a number of MSA- and state-level characteristics as well as information on whether the elections were presidential. Finally, our last model specification includes MSA-specific time trends.
The estimates are robust to the inclusion of additional regressors, especially MSA- and state-level controls and MSA-specific time trends, thus suggesting that time-varying MSA-level characteristics do not significantly alter the registering and voting behavior of U.S. citizens. Focusing on our most complete specification, we find that a 1 standard deviation increase in the intensity of immigration enforcement (about 1.5 times the average level of immigration enforcement) lowers the registration likelihood of U.S. citizens in mixed-status households by 3.6 percentage points (5.3 %)17,18 but leaves registration likelihood unchanged among eligible voters in households without a likely unauthorized immigrant. Because eligible voters residing in mixed-status households are already approximately 4.5 percentage points (12 %) less likely to register than their counterparts in non-mixed-status households, exposure to intensified enforcement only reduces their propensity to register further. Yet, among those registered, immigration enforcement does not alter their voting propensity.
The remaining results in Table 3 are as expected. Older, female, and more-educated eligible voters exhibit a higher propensity to register and vote than their corresponding counterparts. The opposite is, nonetheless, true for Hispanics and single voters. Hispanics are 4 and 3 percentage points (6 % and 4 %) less likely to register and vote, respectively, than their non-Hispanic counterparts, and single respondents are 8 percentage points and 6 percentage points (12 % and 8 %) less likely to register and vote, respectively, than their married equivalents. Finally, as we would expect, the registration and voting likelihood significantly rises during presidential elections by 6 and 21 percentage points (9 % and 28 %), respectively. The fact that the effect of a presidential election is also larger for the likelihood of voting (once registered) compared with only registration is also not surprising.19
In sum, the intensification of immigration enforcement appears to have significantly inhibited the political engagement of eligible voters residing in mixed-status households.20
Using policy leads and placebo tests, we explore whether there were any anticipation effects or preexisting diverging trends (Table S2, Online Resource 1). Our tests confirm that the registration effects from Table 3 cannot be attributed to preexisting trends. We also address the nonrandom location of immigrants and immigration enforcement (Table S3, Online Resource 1). Using a proxy of what the likely exposure to immigration enforcement would have been had immigrants continued to settle in the same areas as their countrymen in 1980, we find that a 1 standard deviation increase in the proxy for immigration enforcement lowers the propensity to register by 3.2 percentage points (4.8 %). The effect is not statistically different from the one in Table 3. As a final identification check, we assess whether past levels of political engagement at the MSA level predict the adoption of tougher enforcement measures (Table S4, Online Resource 1). We estimate the year in which the MSA adopted its first interior enforcement measure as a function of the MSA characteristics prior to the implementation of any enforcement, including the average shares of registered voters and of eligible voters who did participate in the last elections. We find that political engagement prior to the adoption of any interior enforcement measure does not help predict the adoption of tougher enforcement by MSAs. A detailed discussion of these identification checks is in section C, Online Resource 1.
In addition to identification checks, we also perform two robustness checks that examine the consistency of our findings with the use of alternative samples (Table S5, Online Resource 1). When we exclude Maricopa County, an area known for being one of the toughest in immigration enforcement, panel A shows that a 1 standard deviation increase in interior immigration enforcement lowers eligible voters’ propensity to register by 3.6 percentage points (5.4 %). When we exclude states with Trust Acts (areas where enforcement would be more lax), panel B shows that a 1 standard deviation increase in interior immigration enforcement lowers eligible voters’ propensity to register by 3.5 percentage points (5.1 %). These results, which can be found in section D of Online Resource 1, are similar to those in Table 3.
Thus far, we have shown that the intensification of immigration enforcement appears to have curtailed the likelihood of registering to vote among U.S. citizens residing in mixed-status households. Yet, what factors appear responsible for these effects? To address this question, we distinguish between what we refer to as police-based versus employer-based measures. The distinction is important not only because of who is involved in the implementation of the measures but also because of the distinct implications of each group of initiatives. Police-based measures directly involve the local or state police, as in the case of 287(g) agreements, Secure Communities, and OILs. Because the police can stop individuals in the street, on the road, or elsewhere and, if in doubt, request proper identification, these measures involve an element of uncertainty and fear of apprehension of undocumented family members. In contrast, employer-based measures, such as employment verification mandates, are implemented by those hiring and checking the work eligibility of new hires through the electronic E-Verify program. Employees are made aware of the firm’s use of E-Verify, and nonapprovals are not directly associated with deportations.
According to the estimates in Table 4, the effect that intensified immigration enforcement has had on the lower propensity to register of eligible U.S. citizens in mixed-status households is primarily the result of police-based measures, more likely to instill fear among families with unauthorized immigrants trying to evade apprehension and deportation. Specifically, a 1 standard deviation increase in police-based enforcement lowers the propensity of registering to vote among eligible U.S. citizens in mixed-status households by 2.7 percentage points (4.1 %).21,22 The fact that employment-based enforcement has no effect suggests that the potential reasons for the lower registration likelihood among these U.S. citizens are not likely to involve employment opportunities. Perhaps these citizens hesitate to register with a government authority out of fear that their loved ones might be more easily identified.
Having been reassured about the validity, reliability, and sources of the estimated effect of intensified immigration enforcement on the political engagement of eligible voters in mixed-status households, we explore potentially heterogeneous effects by place of birth, ethnicity, gender, and age groups (results are in section E, Online Resource 1). In this manner, we are able to parse the effect of immigration enforcement from that of naturalization or ethnicity as well as assess the incidence of enforcement across demographic groups. We find that U.S.-born eligible voters mostly drive the effects shown in Table 3. A 1 standard deviation increase in enforcement lowers the propensity to register among the U.S.-born by 3.4 percentage points (5 %), with no effect among the naturalized (Table S6, Online Resource 1).
We also know that Hispanics display a lower level of political participation relative to other groups.23 Furthermore, because of their strong historical, cultural, and political roots in the United States, Mexicans in particular might display a differential engagement in the political process.24 As shown in Table S7 (Online Resource 1), a 1 standard deviation increase in immigration enforcement lowers the propensity to register among Hispanics by 2.7 percentage points (5 %). This effect stems from the effect of enforcement on Hispanics of non-Mexican descent: the latter group does not appear to lower their political participation in response to intensified enforcement.25 Rather, among those of Mexican descent who are registered, a 1 standard deviation increase in immigration enforcement raises their propensity to vote by 4.6 percentage points (6.9 %).
Last, we find that most of the effects are concentrated among men and older eligible voters who are not of college age, hinting at the greater engagement of women and the growing number of registration drives on college campuses across the country that induce college-aged U.S. citizens to register (Table S8, Online Resource 1). A 1 standard deviation increase in immigration enforcement lowers men’s propensity to register by 4.2 percentage points (6.4 %) but does not lower women’s likelihood of registering. In addition, a 1 standard deviation increase in enforcement lowers the propensity to register among older age-eligible voters in mixed-status households by 4.8 percentage points (7 %) but not among their younger college-aged counterparts.
Other Forms of Political Engagement
The estimates in Tables 3 and 4 confirm the negative effect that intensified immigration enforcement, through police-based measures, is having on the political engagement of eligible voters residing in mixed-status households by curtailing their propensity to register to vote. This finding leads us to wonder about alternative forms of political engagement, such as volunteering in immigrant or political groups and organizations. Has this demographic group become civically disengaged? Or, rather, has it substituted its electoral participation for a non-electoral form of political participation, such as involvement with the community? After all, volunteering does not require individuals to register with government authorities and, therefore, expose the address of their unauthorized family members. Furthermore, volunteering might strengthen trust in the community and bolster community support and its role as a safety net should the latter be needed, as might be the case in times of intensified immigration enforcement.
To address the aforementioned questions, we turn to data from the 2002–2013 CPS Volunteering Supplements. Respondents in those supplements are asked about (1) their volunteering for any organization or people in their neighborhood (what we refer to as “broad volunteering”); and (2) their volunteering in social and community service groups, civic organizations, immigrant/refugee assistance groups, political parties or advocacy groups, and youth services (which we label “group-specific volunteering”).26 We then reestimate Eq. (1) using broad volunteering and group-specific volunteering as the two dichotomous outcomes. Table 5 displays the estimated effects that intensified immigration enforcement appears to have had on the volunteering behavior of U.S. eligible voters residing in mixed-status households. Although those living in mixed-status households seem less inclined to participate in the volunteering activities measured herein, a 1 standard deviation increase in immigration enforcement consistently raises their propensity to volunteer by 2 percentage points (17 %). When focusing on group-specific volunteering—a significantly more restricted definition of volunteering—we continue to find a positive effect on their propensity to volunteer of approximately 1 percentage point (24 %).
In sum, intensified immigration enforcement does not seem to have curtailed the civic engagement of U.S. citizens aged 18 and older in mixed-status households across the board. Rather, although enforcement appears to have lowered their propensity to register in order to vote, it has not altered the propensity to vote among those registered. In fact, enforcement has strengthened their propensity to volunteer in the community, including in immigrant, political, and advocacy groups. Perhaps these citizens have turned to non-electoral political participation in an effort to assist their families and loved ones, possibly also to ensure community support, in the midst of a proliferating piecemeal approach to immigration enforcement. This finding is consistent with Barreto and Muñoz (2003) and Tran (2016), who found that Latinos are more likely to engage in non-electoral political participation.
Summary and Conclusions
Using data from the 2002–2014 CPS Voting and Registration Supplements, we examine the effect of intensified enforcement on the political engagement of U.S. citizens living in mixed-status households. We find that the toughening of immigration enforcement lowers the propensity to register in order to vote among eligible voters in mixed-status households by 5 %, even though it does not significantly alter the propensity to vote among those already registered. The negative effect of intensified immigration enforcement seems to be working through police-based measures that are more directly linked to the apprehension and deportation of their loved ones, possibly signaling their reticence to register with government authorities and/or their disenchantment with the political process. Heterogeneity tests also reveal how the negative effect of intensified immigration enforcement on the propensity of registering to vote is concentrated among U.S. eligible voters in mixed-status households who are of non-Mexican descent, men, or those not of college age—patterns that are consistent with prior findings on Latino mobilization efforts, political participation, and successful registration drives, correspondingly. Yet, in the face of adversity, U.S. citizens residing in mixed-status households also appear to have strengthened their community engagement, increasing their volunteering activities, including those involving community, immigrant, political and advocacy groups, by a nonnegligible 17 % to 24 %.
In light of the changing profile of the U.S. electorate, which could likely alter presidential, state-, and local-level election results, understanding how the political engagement of eligible voters in mixed-status households is influenced by the piecemeal approach to immigration reform is critical. The demographic shift in the electorate composition and the relevance of immigration policy to this new electorate underscores the importance of understanding the consequences of tougher enforcement as well as the need for targeted registration efforts that ensure their political participation in a democratic society.
See U.S. Department of Homeland Security, Budget in Brief, fiscal years 2003–2013 (http://www.dhs.gov/dhs-budget). Data on apprehensions can be found in Table 33 at http://www.dhs.gov/yearbook-immigration-statistics-2013-enforcement-actions. Data on interior removals can be found at http://www.migrationpolicy.org/research/deportation-and-discretion-reviewing-record-and-options-change.
These are households whose members have different citizenship and immigration statuses. For the purpose of this study, it will include households with, at least, one likely unauthorized immigrant.
DAPA, a proposed program, would temporarily defer the deportation of undocumented parents of U.S. citizens and lawful permanent residents. On June 15, 2017, the Department of Homeland Security Secretary signed a memorandum rescinding the November 20, 2014 memorandum that created the DAPA program.
See section A of Online Resource 1 for a summary of the enforcement policies adopted in the 2000s.
Secure Communities is an information-sharing program used in the apprehension and deportation of unauthorized immigrants. Under the program, local law enforcement agencies can submit information obtained during arrests, such as fingerprints, to an integrated database with ICE that allows for the identification of the immigration status and criminal activity of any individual. See https://www.ice.gov/secure-communities.
The college-age population has been particularly targeted during the past presidential elections, starting with President Obama’s first campaign for the presidency in 2008.
We also experiment with more-restrictive proxies that, in addition to the prior attributes, also require that they be in occupations typically classified as low-skilled—namely, food preparation and serving; building and grounds cleaning and maintenance; personal care and service; sales and related; office and administrative support; farming; fishing; forestry; construction; extraction; installation, maintenance, and repair; and production. Results can be found in Online Resource 1, Table S1.
Data restrictions inhibit us from properly implementing more sophisticated statistical imputation methods. As Van Hook et al. (2015) noted, the precision of the estimates derived using statistical imputation methods depends on (1) the joint observation of every pair of variables and (2) both target and donor samples being drawn from the same universe. In our case, where unauthorized status is not observed in the CPS supplement (target data set), we would need to have information on the civic outcomes in a donor sample that has information on the unauthorized status and originates from the same sample universe. In addition, the donor sample would have to be representative of smaller geographic units, such as MSAs in our case, and informative over the 2002–2014 period. It is unclear whether such a data set exists.
The CPS allows us to group household members using a unique household identification number. Using our definition for likely unauthorized migrants, we are able to identify how many members of the household are likely unauthorized. We then classify the household as mixed-status if the household has at least one likely unauthorized member. This approach reveals only whether a likely unauthorized immigrant resides in the same household as a U.S. citizen. It does not allow us to identify the relationship of the unauthorized immigrant to the U.S. citizen. To gain insight into relationship status of these unauthorized migrants, we also followed a second approach, which is to use parent location and spouse location in the CPS to link U.S. citizens to unauthorized parents or spouses. We then compare the number of U.S. citizens living with an unauthorized parent or spouse to the number of U.S. citizens living in a mixed-family household to find that the unauthorized household members are predominantly parents or spouses.
The NCSL is a bipartisan nongovernmental organization that serves the members and staff of state legislatures of the United States. Information on immigration can be found online (http://www.ncsl.org/research/immigration.aspx).
, where k = SC, MSA 287(g), State 287(g), OIL, and E-verify.
The index is a proxy of the intensity of immigration enforcement to which respondents in a particular MSA might be exposed. The true intensity of any enforcement measure will inevitably vary across jurisdictions because each one is different and might implement alike measures more or less strictly depending on who is in charge of its implementation or other unobserved local traits. To address that limitation, we include area fixed effects as well as area-specific time trends intended to capture such idiosyncrasies.
In the heterogeneity analysis, our five individual indices are separated into police- and employment-based measures. For or k = SC, MSA 287(g), State 287(g), and OIL. It is the sum of the four measures for each MSA for each year. The .
Many of the immigration enforcement measures, such as the Secure Communities and the 287(g) agreements, were designed to substitute, replace, or continue one another. In addition, they overlap, which has the potential to exponentially raise their effect given that each measure builds on the police infrastructure established by the other.
Because ethnicity (in particular, being Hispanic) is used as a predictive trait of being unauthorized, most eligible voters in households with a likely unauthorized immigrant are going to be of the same ethnicity. Exceptions would include, for example, non-Hispanic U.S. citizens married to a likely unauthorized immigrant.
An example would be the case with confounding changes in wealth among mixed-status households if they suffered greater income losses during the recession.
As shown in Table 1, the average level of enforcement is 0.790 with a standard deviation of 0.911. Additionally, the mean for registration among households is 0.669. The estimated effect is computed as (1 SD increase in enforcement × coefficient) / mean of the dependent variable. That is, (0.911 × 0.039) = 0.0355, or 3.6 percentage points. Or, if we want to use a unit-less measure, we can divide 0.0355 by 0.669, which gives us 0.0531, or 5.3 %.
To serve as a reference, this is a similar effect to that of being Hispanic—a trait that is associated with a 6 % lower likelihood of registering to vote, as we shall discuss later herein.
In separate analyses, we assess whether the registration and voting likelihood of eligible voters in mixed-status households who were exposed to intensified immigration enforcement significantly varies during a presidential election year. We do not find any statistically significant evidence of that being the case. Results are available from the authors.
Our finding proves robust to the use of an alternative definition of likely unauthorized that restricts that denomination to individuals who, in addition to being Hispanic, being noncitizens, having less than a high school diploma, and having spent five or more years in the United States, are employed in low-skill occupations. See Table S1 in Online Resource 1.
The standard deviation of police-based immigration enforcement is 0.669. Thus, we interpret the effect as 0.041 × 0.669 = 0.0274, or 2.7 percentage points; we then divide by the dependent variable mean, which is 0.669, to get 4.1 %.
Separate analyses reveal that most of this impact originates from the Secure Communities program. However, for the reasons noted earlier, that effect should be interpreted with caution. Results are available from the authors.
This may be more the case for Mexicans residing in California where there are stronger networks and political mobilization efforts among Mexicans in the state (Pantoja et al. 2001).
Albeit not statistically different from 0 at conventional levels, the coefficient on the interaction term in the registration equation for Hispanics of Mexican descent is close to being significant with a t statistic equal to 1.56.
We group these categories owing to the relative low incidence of volunteering in some of them. Using categories allows us to identify more specific ways (including more time-intensive) ways that people volunteer.