Immigrant Status, Citizenship, and Victimization Risk in the United States: New Findings from the National Crime Victimization Survey (NCVS)

Abstract

Until recently, national-level data on criminal victimization in the United States did not include information on immigrant or citizenship status of respondents. This data-infrastructure limitation has hindered scientific understanding of whether immigrants are more or less likely than native-born Americans to be criminally victimized and how victimization may vary among immigrants of different statuses. We address these issues in the present study by using new data from the 2017–2018 National Crime Victimization Survey (NCVS) to explore the association between citizenship status and victimization risk in a nationally representative sample of households and persons aged 12 years and older. The research is guided by a theoretical framing that integrates insights from studies of citizenship with the literature on immigration and crime, as well as theories of victimization. We find that a person’s foreign-born status (but not their acquired U.S. citizenship) confers protection against victimization. We also find that the protective benefit associated with being foreign-born does not extend to those with ambiguous citizenship status, who in our data exhibit attributes similar to the known characteristics of undocumented immigrants. We conclude by discussing the implications of our findings and the potential ways to extend the research.

Does citizenship status influence one’s risk of criminal victimization? Despite a long tradition of research on the association between immigration and crime (e.g., Hagan & Palloni, 1998; Lyons, Vélez, & Santoro, 2013; Martinez & Lee, 2000; Ousey & Kubrin, 2018; Park, 1928;Thomas & Znaniecki, 1927; Xie & Baumer, 2018), and despite research attention devoted to understanding citizenship and immigrant integration (e.g., Brubaker, 1992; Massey & Bartley, 2005; Portes & Rumbaut, 2006), the answer to this question remains uncertain. This uncertainty has persisted largely because the major national data sources of crime in the U.S. have long failed to collect information about immigration status (Ousey & Kubrin, 2018; Zatz & Smith, 2012). In recent decades, this important knowledge gap has been amplified by the scale of transnational migratory flows. The number of foreign-born residents in the U.S. — approximately 45 million — has reached historical highs, with the group now accounting for nearly 14 percent of the total population (Migration Policy Institute, 2020). Additionally, the foreign-born population has become increasingly heterogeneous with respect to their citizenship status, with recent estimates showing that the group now comprises large shares of naturalized citizens (45%), documented noncitizens (32%), and undocumented noncitizens (23%; see Pew Research Center, 2020).

The present study advances knowledge about immigration and crime by examining the relationship between citizenship status and victimization risk with new data from the U.S. National Crime Victimization Survey (NCVS) conducted in 2017–2018. The new NCVS data are unique due to the addition of a question on respondents’ citizenship status. Following other national surveys, such as the American Community Survey (ACS) and the Current Population Survey (CPS), the NCVS now asks respondents to self-identify as U.S.-born citizens, naturalized citizens, and non-U.S. citizens. The question purposely appears at the end of the survey, after respondents have completed all other parts of the interview (Bureau of Justice Statistics, 2019). This is an important feature of the survey design that should enhance the validity of efforts to compare the victimization experiences of those who self-identify as immigrants (i.e., naturalized citizens or noncitizens) and those who identify as U.S.-born citizens (Dillman, 2007). In addition, although the NCVS does not ask for the documentation status of those who identify as noncitizens¹, there is another feature of the NCVS data collection that makes the data potentially valuable for providing insights about victimization among noncitizens who may be undocumented immigrants. Specifically, those who refused to answer the citizenship question or responded “don’t know” exhibit attributes similar not only to those observed for the noncitizen immigrant population, but also to the characteristics of the undocumented immigrant population. Although it is not possible to definitively equate these “ambiguous citizenship” responses to a particular citizenship status, the presence of this group in the data offers an opportunity to broaden the study of immigration and crime in a potentially important way.

Using panel data from the 2017–2018 NCVS, our analysis considers how people of four different citizenship statuses — U.S.-born citizens, naturalized citizens, noncitizens, and those with ambiguous citizenship status — compare with respect to violent and property victimization, net of other household and personal characteristics the survey measures. To our knowledge, the study offers the first multivariable analysis of citizenship status and victimization risk in the U.S. Furthermore, it incorporates nonresponse on the NCVS citizenship question in a novel fashion to explore patterns of victimization among a group of noncitizens who, in many respects, parallel the undocumented immigrant population. This offers both conceptual and methodological innovations to the literature on immigration and crime, and it extends research that has, in the past, relied primarily on field interviews obtained from accessible communities to learn about the experiences of immigrants (e.g., Bucher, Manasse, & Tarasawa, 2010; Fussell, 2011; Grubb & Bouffard, 2014; Menjívar & Abrego, 2012).

We begin by illuminating reasons why there may be important variations in victimization risk between immigrants and U.S.-born citizens and among immigrants of different citizenship statuses. We do so by integrating theories of victimization risk with insights from the interdisciplinary literatures on citizenship and on immigration and crime. This synthesis reviews the context in which citizenship data are collected and how the meanings of citizenship should be viewed and understood in today’s U.S. society. We argue that the acquisition of citizenship (or lack thereof) is associated with heterogeneity among the foreign-born population in sociodemographic attributes and their capacity to maintain social ties, establish connections to social institutions, and access legal remedies, all of which may affect victimization risk. We then describe the NCVS data used in the study and the methods we apply to analyze how citizenship status is related to victimization risk. After reporting the results of our investigation, we underscore how the findings provide important insights during an era in which the nation and its criminal justice and immigration systems are facing major challenges amid tension surrounding immigration and growing concerns about social and racial-ethnic injustice.

1. BACKGROUND

1.1. A Critical Shortage of Citizenship Data in the Study of Crime and Victimization

In the U.S., collecting citizenship data is a challenging task, and the public debate that eventually culminated in the 2019 U.S. Supreme Court ruling to bar a citizenship question from the 2020 census (Department of Commerce v. New York, 2019) is an instructive example showing how complex and controversial — both methodologically and socially — the matter can be. In the census debate, many observers emphasized the importance of a complete count of the U.S. population, regardless of one’s citizenship status, because the population count is the basis for political representation and funding allocations (Sullivan, 2020). The eventual decision to reject the citizenship question for the census reflects the well-founded concern that it may reduce public trust and hurt the interests of local communities, especially the interests of racial-ethnic minorities (Brown, Heggeness, Dorinski, Warren,& Yi, 2019; Van Hook, 2018).

Like the census data, many of the largest data collections in the U.S. on crime (e.g., the Uniform Crime Reporting [UCR] program, the National Vital Statistics System [NVSS], and the NCVS prior to 2017) do not record any immigration information (Lauritsen & Cork, 2018; Ousey & Kubrin, 2018; Xie & Baumer, 2018). The collection of immigration data was not a high priority for government agencies overseeing the initial development of these data systems (i.e., from the late 1920s to the early 1970s, when the U.S. immigration rate was trending down to a record low level). These data shortcomings were, and still are, difficult to overcome, even as the U.S. has experienced a large influx of immigrants during the past several decades.

Because it was impossible to identify individual victims by citizenship status on a national scale until the recent release of the NCVS data in 2017 (Morgan & Truman, 2018), theory development in this area has been limited. Much of what we know has been derived from legal and policy studies, qualitative research, special surveys that examine some selected groups of the hard-to-reach immigrant population, or from ecological studies that make area-level comparisons across cities or other geographic units. These studies have been valuable, as they have challenged a monolithic view of immigrants (for a review, see Zatz & Smith, 2012). Yet, it is also important to expand on these efforts. We do so by integrating insights from these studies with the theoretical literature on victimization to delineate hypotheses for how victimization risk may vary across different citizenship statuses (U.S.-born citizen, naturalized citizen, noncitizen, ambiguous citizenship status).²

1.2. Theoretical Expectations and Hypotheses

Although the prevailing research on immigration and crime does not explicitly use the concept of citizenship, this literature offers theoretical insights about how victimization risk may be expected to differ between the U.S. natives (who are citizens by birth) and the foreign born as a whole. We begin by describing relevant arguments from this literature, which as we summarize in Figure 1, leads us to expect the foreign born to experience lower levels of victimization compared to native-born citizens. We then differentiate the foreign-born population by citizenship status, drawing from literatures on migration experiences, citizenship, and the etiology of crime and victimization to describe how citizenship status may affect victimization risk in the U.S. As shown in Figure 1, this synthesis leads us to anticipate that naturalized citizens will have a lower risk of victimization compared to noncitizens, and that respondents with ambiguous citizenship status (who exhibit attributes that mirror some of the known characteristics of the undocumented immigrant population) will experience a higher risk of victimization than noncitizens and naturalized citizens.

Figure 1.

Summary of citizenship status categories and hypotheses

H1: The foreign born as a whole may experience lower levels of victimization compared to native-born citizens, net of the effects of other personal and household characteristics.

H2: Naturalized citizens may have a lower risk of victimization than their noncitizen peers, net of the effects of other personal and household characteristics.

H3: The ambiguous-status group identified in our study may experience a higher estimated risk of victimization compared to respondents who self-identify as naturalized citizens or noncitizens, net of the effects of other personal and household characteristics.

1.2.1. How victimization risk may differ between the foreign born and U.S. natives

Socioeconomic and Demographic Differences

For the foreign-born population as a whole, national statistics reveal lower levels of educational attainment, lower earnings, and higher rates of poverty than native-born residents (Pew Research Center, 2020; U.S. Census Bureau, n.d.). These socioeconomic characteristics have been found in previous research to be some of the major structural and opportunity-related predictors of elevated victimization risk (Sampson & Lauritsen, 1994). Yet, as many migration/immigration scholars have pointed out, the foreign-born population also possesses many attributes that often serve to offset economic and educational disadvantages and promote well-being, which may limit their exposure to adverse social, behavioral, and health outcomes (National Academies of Science, Engineering, & Medicine, 2018). Perhaps most pertinent to victimization risk, compared to the native-born population, foreign-born persons are somewhat older on average, are more likely to be employed and to work more hours, and are more likely to be married and live in a family household (Pew Research Center, 2020; U.S. Census Bureau, n.d.). These characteristics may serve to keep immigrants away from “high-risk times, places, and people,” which may lower their victimization risk as lifestyle and routine activities theories have suggested, from the point of view that victimization depends on the convergence of a motivated offender, a suitable victim, and the absence of capable guardianship (Hindelang, Gottfredson, & Garofalo, 1978: 245; also see Cohen & Felson, 1979; Lauritsen & Rezey, 2018). Beyond these observed patterns, the literature has highlighted two mechanisms that support the expectation of lower levels of victimization among the foreign born when compared to U.S. natives, net of the observed socioeconomic and demographic attributes: selective migration and community externalities.

Selective Migration

Through statistical and field research, migration scholars have documented that there is significant positive selection involved in the process that determines who emigrates to and remains in America, widely referenced as a “healthy migrant effect” (see Gubernskaya, Bean, & Van Hook, 2013; Palloni & Arias, 2004). The overarching logic of this perspective is that both migration to the U.S., and the circumstances and decisions that determine whether a migrant remains in the country, tend to be positively correlated with several attributes of migrants that are harder to measure than variables typically captured by demographic surveys, such as commitment to family and socialbonds, relatively good physical and mental health, and low levels of involvement in risky behaviors that reduce their contact with potential offenders (Cunningham, Ruben, & Narayan, 2008; National Academies of Science, Engineering, & Medicine, 2018). These selection processes result in a foreign-born population that is less prone to victimization than U.S.-born citizens. In particular, research situated within the lifestyle-routine activity theoretical perspective indicates that a variety of risky behaviors that promote criminal victimization are less prevalent among the foreign born, including alcohol and drug dependence (Salas-Wright, Vaughn, Goings, Córdova, & Schwartz, 2018; Szaflarski, Cubbins, & Ying, 2011), antisocial conduct (Vaughn et al., 2014), and criminal offending (e.g., Bersani & Piquero, 2017; MacDonald & Sampson, 2012; Martinez & Lee, 2000).

Community Externalities

Community-based research has also indicated that immigrants tend to settle in extended family and community settings that exhibit “protective” social and cultural features (National Academies of Science, Engineering, & Medicine, 2018; Riosmena, Wong, & Palloni, 2013), which have been linked to higher levels of guardianship and informal social control that may reduce victimization (Sampson, 2008). Contemporary theoretical arguments suggest that the concentration of immigrants in a community helps promote economic and social revitalization, yielding not only opportunities for labor market attachment, but also social supports for promoting family stability and social networks (Kubrin & Desmond, 2015; Portes & Rumbaut, 2006). A growing empirical literature connects these conditions to lower levels of crime and victimization within immigrant communities (e.g., Lee & Martinez, 2002; Lyons, Vélez, & Santoro, 2013; Painter-Davis, 2016; Vélez, 2009; Xie & Baumer, 2018).

These individual- and community-level mechanisms — most notably less involvement in risky behaviors and greater exposure to “protective” immigrant community contexts — highlight why we anticipate lower levels of victimization among the foreign born as a whole, compared to native-born citizens, net of the impact of demographic and socioeconomic characteristics measured in the NCVS (Hypothesis 1, Figure 1). Although the data used for our study do not permit a comprehensive assessment of the validity of the suggested mechanisms, they allow a valuable comparison at the national level for the first time that reveals how U.S.-born citizens and the foreign-born population compare with regard to violent and property victimizations. The comparison examined in the present study extends existing reports of violent victimization rates calculated by the BJS (Morgan & Oudekerk, 2019; Morgan & Truman, 2018) by adding information on property victimizations. Additionally, our assessment extends the relatively simple (bivariate) rate calculation reported by the BJS to a multivariable framework and new findings.

1.2.2. How citizenship status may affect victimization risk among the foreign born

The heterogeneity of the foreign-born population and their different life experiences motivate us to disaggregate this population by citizenship status and explore possible heterogeneity with regard to victimization. Based on extant theoretical and empirical literature (Bloemraad, 2018; Bosniak, 1988; Menjívar & Abrego, 2012), we suggest that citizenship or its absence is a distinctive form of stratification that could meaningfully influence victimization. To delineate specific hypotheses, we integrate victimization theories with insights about how national and international forces have coalesced to shape the flows of documented and undocumented immigrants to the U.S. over the past several decades (Massey, 1999; Menjívar & Perreira, 2019). We suggest that differences in victimization may be a result of the different life conditions that lead some, but not all, immigrants to become naturalized citizens, and we also link victimization risk to the structure of U.S. immigration laws and enforcement practices aimed at controlling immigrants (Johnson, 1996/1997; Varsani, 2010). This synthesis reveals that citizenship status is a potentially important symbol associated with different attributes and life experiences that may influence victimization through a plurality of pathways. We highlight three primary themes in this discussion, each of which implies a status hierarchy among the foreign born that predicts the following rank order of victimization risk among immigrants, listed from lowest to highest: naturalized citizens, followed by noncitizens with documentation, and then undocumented noncitizens (who are more likely to be concentrated in the “ambiguous citizenship status” respondents in the NCVS).

Citizenship Status and Stratification by Demographics and Socioeconomic Factors

Naturalized citizens, documented noncitizens, and the undocumented exhibit differences in a variety of demographic and socioeconomic attributes that are associated with victimization. Such differences exist, as research on immigration indicates, because there are different selections for migration (e.g., migration for family, economic, humanitarian, or other reasons) that indicate immigrants’ varied access to human capital (Van Hook, Brown, & Bean, 2006), and because immigrants go through different experiences of incorporation in the U.S. (Portes & Rumbaut, 2006). Data from the ACS and CPS indicate that, compared to non-citizens, foreign-born persons who are naturalized citizens in the U.S. tend to be older, are more likely to be married, and have higher socioeconomic status (Sumption & Flamm, 2012; U.S. Census Bureau, n.d.). Each of these attributes is posited to lower victimization risk through beneficial effects on lifestyles (Hindelang, Gottfredson, & Garofalo, 1978; Miethe, Stafford, & Long, 1987). In contrast, available data indicate that undocumented immigrants tend to be younger, less educated, more likely to be renters, and less apt to be married (Gelatt & Zong, 2018), all of which have been connected to routine activities and lifestyles that increase crime exposure (Cohen & Felson, 1979; Cohen, Kluegel, & Land, 1981). Additionally, compared to other immigrants, the undocumented tend to arrive in the U.S. under conditions of extreme economic deprivation and to experience ongoing financial strain while in the country (Brown & Bean, 2006), a structural feature also emphasized as conductive to victimization in both lifestyle-routine activities theories and structural-choice theories of victimization (see Meier & Miethe, 1993). For example, the Mexican debt crisis in the 1980s and the more recent Latin American migration crisis, which was brought by drought, poverty, violence, and hurricanes, compelled large numbers of migrants to travel north without proper authorization (Isacson, Meyer, & Morales, 2014; Menjívar & Perreira, 2019). Some migrants entered the U.S. legally but then overstayed their visas for fear of being unable to meet unforeseen future migration requirements (Cleaveland, 2010). Unauthorized immigrants from other regions of the world, such as Asia, Europe, and Africa, have faced similar migration circumstances (e.g., Chan, 2013). Hence, compared to naturalized citizens or noncitizens with proper documentation, the undocumented are more likely to migrate under conditions of significant economic disadvantage, which is often amplified by the debt they frequently must absorb to pursue options of border crossings or illegal stay and by ongoing labor market constraints associated with their status (Rivera-Batiz, 1999; Yoshikawa, 2011; Zhang, 2007). The empirical research has long shown that persons who experience extreme economic distress are at greater risk for victimization (Harrell, Langton, Berzofsky, Couzens, & Smiley-McDonald, 2014; Sampson & Lauritsent, 1994).

Citizenship Status and Implications for Life Circumstances in the U.S.

Net of differences in demographic and socioeconomic attributes, the citizenship (and documentation) status of the foreign born often has significant, albeit divergent, impact on immigrants’ daily lives, which in turn may translate into variation in victimization risk. On the one hand, attaining citizenship through the naturalization process often yields positive benefits for the foreign born (Bloemraad, 2018) that may reduce victimization risk. Under prevailing laws, for example, the standard benefits of citizenship include increased access to a much wider range of employment opportunities, the right to vote, and opportunities for family reunification (Aptekar, 2016). Thus, attaining citizenship can strengthen social ties and connections to core social institutions, which have been identified as important for providing social support and social controls that reduce victimization risk (Policastro & Daigle, 2019) and exposure to high levels of crime within one’s community (e.g., Messner & Rosenfeld, 1994).

In contrast, noncitizens, and especially the undocumented, often face unique challenges of isolation or exclusion that sever family and kinship ties and limit access to social supports that may protect against victimization (Menjívar & Abrego, 2012). Contemporary legislation has placed broad restrictions on legal immigrants’ access to public benefits through measures such as the 1996 Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) and the 1996 Illegal Immigration Reform and Immigrant Responsibility Act (IIRIRA), which made many legal immigrants ineligible for federally funded benefits (Ewing, Martinez, & Rumbaut, 2015; Van Hook, Brown, & Bean, 2006). More recently, after the terrorist attacks of 2001, the documentation status of the foreign born has increasingly become a barrier that separates them from the mainstream society, as the nation passed more restrictive employment eligibility verification laws and expanded the use of detention and deportation as a form of social control (Forrester & Nowrasteh, 2018; Kubrin, 2014; Miles & Cox, 2014; Treyger, Chalfin, & Loeffler, 2014). As other scholars have noted, the persistent threat of deportation adversely affects immigrants and challenges their ability to adjust to life in the U.S. (Dreby, 2015;). Such effects have been reported in housing, employment, health, and other life domains, which put noncitizen immigrants in a vulnerable social position (Zatz & Smith, 2012). This is particularly the case for the undocumented, who often face the greatest risk of disruption to family, neighborhood, and other “protective” social networks, which in turn could increase their risk of victimization (Policastro & Daigle, 2019).

Citizenship Status and Differential Access to the Law

As other scholars have pointed out, there is also considerable heterogeneity among the foreign born in perceived access to the law that may influence victimization risk (e.g., Davis & Hendricks, 2007; Menjívar & Abrego, 2012). This is another aspect of immigrant life that is closely linked to the themes discussed in the previous section, but it is worth a separate discussion because of the large media and scholarly attention given to this issue (e.g., Engelbrecht, 2018; Medina, 2017; Menjívar & Bejarano, 2004). Attaining citizenship through the naturalization process enables foreign-born persons to access legal remedies without fear of deportation. In contrast, noncitizens may be reluctant to seek legal remedies, such as contacting the police to resolve interpersonal disputes, due to concerns that police contact may lead to deportation for them or their family and friends (Zatz & Smith, 2012). This tendency to avoid enlisting the help of the law is likely to be most acute among the undocumented (Chapin, 2011). Although not well addressed in the theoretical literature on victimization, a synthesis of research findings from interviews with law enforcement officials (e.g., Winton & Yi, 2006), analyses of written laws and enforcement policies (e.g., Kittrie, 2005), and in-depth interviews with immigrant victims (e.g., Menjívar & Abrego, 2012) suggests that both criminal justice professionals and members of immigrant communities believe that different treatment by the law may undermine public trust in the police and increase victimization risk as a result. Furthermore, research on offender target selection suggests that individuals who are perceived as less likely to contact the police may be especially vulnerable to criminal victimization because this mitigates risk for offenders (Wright & Decker, 1997). There is growing evidence in the literature to support this argument, which showed that people who do not report incidents to the police are more likely to experience subsequent victimizations (Felson, Ackerman, & Gallagher, 2005; Ranapurwala, Berg, & Casteel, 2016; Xie & Lynch, 2017).

In summary, based on the literature reviewed, we hypothesize that among the foreign-born respondents in our study, naturalized citizens may have a lower risk of victimization than their noncitizen peers, net of the effects of other personal and household characteristics (Hypothesis 2, Figure 1), given the economic, social and political benefits of citizenship. Additionally, we hypothesize that the ambiguous-status group identified in our study may experience a higher estimated risk of victimization compared to respondents who self-identify as naturalized citizens or noncitizens, net of the effects of other personal and household characteristics (Hypothesis 3, Figure 1). We suspect this because the ambiguous status group is likely to contain a higher proportion of undocumented immigrants.³ As we conveyed above, these multivariable expectations are due to immigrant heterogeneity in the capacity to maintain social ties, establish connections to social institutions, and access legal remedies that are often pivotal for protecting people from crime.

Thus far, these predictions have not been examined in a multivariable setting, but a growing number of studies have considered relevant immigrant experiences in the U.S., including Mexican and Central American migrants (Menjívar & Abrego, 2012), Latino day laborers (Negi, Cepeda, & Valdez, 2013), janitors (Cranford, 2005), Latino undocumented women (Salcido & Adelman, 2004), Latino undocumented men (Bucher, Manasse, & Tarasawa, 2010), Arab immigrants (Hendricks, Ortiz, Sugie, & Miller, 2007), and Asian immigrants (Grubb & Bouffard, 2014), often recruited through local employers, work sites, churches, and community-based organizations (for additional studies, see reviews by McDonald & Erez, 2007; Zatz & Smith, 2012; Pendergast, Wadsworth, & LePree, 2018). Respondents in these studies frequently reported experiencing crimes ranging from verbal and physical violence to property-related crimes such as wage exploitation, theft, robbery, and burglary. These studies offer important clues as to how lacking citizenship or formal status may lead immigrants to situations that increase their victimization risk. These high-risk situations include work in low-wage employment sectors with few benefits and little job security (Cranford, 2005), cash-based transactions (Fussell, 2011), work and residence in transient and crowded environments (Abrego, 2006), being reluctant to contact the police for fear of government attention (Bucher, Manasse, & Tarasawa, 2010), and lacking information about or access to social services (Zadnik, Sabina, & Cuevas, 2016). Individuals possessing these characteristics are often considered attractive targets for crime, whose target suitability is increased by their physical visibility (e.g., race/ethnicity, language accent), potential yield (e.g., their tendency to rely on cash exchanges), greater exposure to motivated offenders (e.g., presence in impoverished and urban neighborhoods), and a lack of guardianship given their strained relationship with law enforcement (Menjívar & Abrego, 2012). In short, previous studies of immigrants have highlighted the vulnerability of noncitizens as victims of crime, especially among those without documentation. The question remains whether, and to what extent, those vulnerabilities as described in these studies may produce a measurable difference in victimization outcomes in more generalizable samples. The large, nationally representative sample of the NCVS provides the data needed to expand knowledge about this question.

2. DATA AND METHODS

2.1. NCVS 2017–2018 Data

The NCVS is the nation’s largest survey on criminal victimization. The Census Bureau conducts the survey for the Bureau of Justice Statistics (BJS) to provide national estimates of violent and property crime victimization for the population of non-institutionalized residents aged 12 years and older (Lynch & Addington, 2007). In addition to crime incidents, it also collects information on factors that may influence victimization risk, including characteristics of individuals and households (Lauritsen & Rezey, 2018).

The NCVS uses a rotating panel design in which sampled households and eligible respondents within the households are scheduled to be interviewed every six months over a three-year period, with each respondent contributing up to seven interviews. Since 2015, nationally representative samples have been selected using a stratified multistage cluster sampling method based on the 2010 census population frame. In 2017–2018, 482,469 persons and 296,523 households were interviewed by the NCVS and the response rates were 83% (persons) and 74% (households), respectively. These response rates were comparable to those of the year 2016 data collection (84% and 78%), before the citizenship question was added.⁴

NCVS respondents are asked to recall incidents of violence (i.e., rape or sexual assault, robbery, aggravated and simple assault) and household property crimes (i.e., burglary, motor vehicle theft, and other types of theft) during a window, or reference period, of the preceding six months. Approximately 56% of violent victimization and 65% of property victimization detected in the survey were not reported to the police, which indicates the importance of the data for assessing victimizations among citizens and noncitizens who may have reasons to be suspicious of the police and would be reluctant to rely on the police to address their crime issues (Ousey & Kubrin, 2018; Zatz & Smith, 2012; Xie & Baumer, 2019).

2.2. Measures

2.2.1. Violent and property victimizations

For the descriptive results presented below, we estimated violent and property victimization rates for each citizenship-status group using the BJS recommended procedure of counting the actual number of victimizations reported by the victims, up to a maximum of 10, to minimize the effect of extreme outliers on rates (Lauritsen, Owens, Planty, Rand, & Truman, 2012). For the multivariable analyses, we adopt a different approach by measuring victimization using dichotomous variables rather than the continuous victimization rates since it is rare for respondents to indicate multiple incidents of violent or property crimes during a reference period. We include a dichotomous outcome for violent victimization, which captures whether respondents experienced one or more incidents of rape, sexual assault, robbery, aggravated assault, or simple assault during the preceding six months (1=yes, 0=no), and a dichotomous outcome for property victimization, which captures whether the household has experienced one or more incidents of burglary, motor vehicle theft, or other theft (1=yes, 0=no).

2.2.2. Citizenship status

The original NCVS codes for the citizenship item distinguish U.S.-born citizens (i.e., persons born in the U.S., U.S. territory, or abroad to U.S. parents), naturalized citizens, noncitizens, and a “residue” value for those who refused to answer or answered “don’t know” to the question. We kept the first three codes as recorded in the data, but we took the following steps to integrate data for the “residue” group in a more conceptually meaningful way. First, using person identifiers in the 2017–2018 NCVS files, we matched observations of the same persons across interviews to identify instances in which a person changed responses to the citizenship question from “residue” to “citizen”.⁵ When this occurred, the observation was excluded from the “residue” group. Second, using household identifiers, we linked individuals to others in the same household, and those in the following categories were removed from the “residue” group because it is likely that they are either citizens or have a legal pathway to citizenship: respondents whose parents are U.S.-born citizens, respondents who are minor children (under the age of 18) of naturalized citizens, and respondents who are spouses of U.S.-born or naturalized citizens.⁶ Third, because the 1986 Immigration Reform and Control Act (IRCA) legalized most unauthorized immigrants who had arrived in the country prior to 1982 (Warren, 2014), those who have been residing in the sampled units since at least 1982 were removed from the “residue” group. Finally, because unauthorized immigrants are not eligible to enlist in the armed forces, respondents who are, or once were, armed forces members were excluded from the “residue” group. In each of these cases, the criterion suggests a high probability that the respondents are legal residents, but this is not a certainty and it remains unclear which of the authorized citizenship groups to which they likely belong (i.e., native-born or naturalized citizen). In light of the uncertainty, we excluded them from the analyses.⁷

Overall, the final person-level sample for the analyses of violent victimization included interviews of 422,500 U.S.-born citizens, 33,728 naturalized citizens, 23,533 noncitizens, and 1,393 with an ambiguous citizenship status (i.e., those who refused to answer the question or provided a “don’t know” response). Using the citizenship status of each household’s reference person (i.e., the person owning, buying, or renting the home, henceforth called the household head), the sample for property crime included 250,404 households headed by U.S-born citizens, 19,365 households headed by naturalized citizens, 12,018 households headed by noncitizens, and 805 households headed by persons with an ambiguous citizenship status.⁸

2.2.3. Control variables

We controlled for previously identified demographic and socioeconomic factors of criminal victimization including respondents’ race/ethnicity (Latino, Asian, White, Black, or other non-Latino race), age, sex, marital status, years of education, employment status (employed or not), household income (1–14 scale), household size, homeownership status (owned or rented), and years of residence (Dugan & Apel, 2003; Lauritsen & Rezey, 2018; Schreck & Fisher, 2004). A binary indicator of whether the home is located in a central city neighborhood and four census regional variables (South, West, Midwest, and Northeast) were added to adjust for area differences in crime (McDowall & Loftin, 2009) and residential patterns associated with citizenship status (Iceland, 2009). As evidence suggests that repeated interviews may lead to reduced frequencies of victimization in the NCVS because of either testing effects or respondent fatigue (Hart, Rennison, & Gibson, 2005; Lynch, Berbaum, & Planty, 2002), we also included time in sample (number of survey interviews completed by each respondent) to allow for such effects.

2.3. Statistical Analyses

We begin the analyses with a comparison of demographic and socioeconomic attributes across the four citizenship status categories examined (U.S.-born citizens, naturalized citizens, noncitizens, and ambiguous citizenship status). We then show how the NCVS ambiguous status group compares to the known statistical profile of undocumented immigrants in the U.S. based on available research. After presenting these descriptive results, we test our hypotheses about how citizenship status may impact victimization risk using logistic regression models. The model coefficients and variance estimates for these models were calculated using survey design variables (pseudo-strata and half sample codes) and sample weights to account for the stratified multistage cluster design of the NCVS.⁹ We tested the models for multicollinearity, and a review of collinearity diagnostics (bivariate correlations, variance inflation factors, and condition number tests) indicated that there were no significant multicollinearity problems in the models.

3. RESULTS

3.1. Group Differences in Demographic and Socioeconomic Characteristics

We use two tables to show the demographic and socioeconomic characteristics of the four groups being examined, as reflected in the NCVS data. Table 1 shows descriptive statistics for the sample of persons aged 12 and older who were surveyed about experiences with violence, while Table 2 presents parallel information for the sample of household heads surveyed about property crimes. The tables reveal notable differences between U.S.-born citizens and the other three groups for almost all variables listed (i.e., race/ethnicity, age, marital status, education, employment status, household income, homeownership status, years of residence, and residential location) that may contribute to differences in victimization risk.¹⁰

Table 1.

Characteristics of residents aged 12 and older by citizenship status, 2017–2018

	U.S.-born citizen		Naturalized citizen		Noncitizen		Refused or don’t know
Characteristics	Mean	(SD)	Mean	(SD)	Mean	(SD)	Mean	(SD)
Latino	.12	(.30)	.37	(.41)	.54	(.42)	.58	(.39)
Asian, non-Latino	.02	(.14)	.32	(.39)	.26	(.37)	.22	(.33)
White, non-Latino	.70	(.42)	.20	(.34)	.12	(.27)	.13	(.27)
Black, non-Latino	.13	(.31)	.09	(.24)	.06	(.20)	.05	(.17)
Other race, non-Latino	.02	(.14)	.02	(.11)	.01	(.09)	.02	(.10)
Age	44.27	(18.77)	50.68	(14.35)	39.00	(12.27)	40.53	(12.02)
Male	.49	(.46)	.47	(.42)	.51	(.42)	.51	(.40)
Married	.45	(.46)	.66	(.40)	.57	(.41)	.58	(.39)
Divorced	.11	(.28)	.08	(.23)	.05	(.17)	.05	(.17)
Separated	.02	(.12)	.03	(.14)	.03	(.15)	.03	(.14)
Widowed	.06	(.21)	.06	(.20)	.03	(.13)	.02	(.10)
Never married	.37	(.45)	.18	(.32)	.32	(.39)	.32	(.37)
Education in years	13.20	(2.83)	13.16	(3.31)	11.90	(3.73)	11.87	(3.13)
Employed	.61	(.45)	.61	(.41)	.65	(.40)	.67	(.38)
Household income	12.12	(3.84)	12.14	(3.37)	10.82	(3.61)	10.79	(3.40)
Household size	2.76	(1.38)	3.08	(1.34)	3.45	(1.46)	3.05	(1.32)
Homeowner	.68	(.43)	.66	(.40)	.34	(.40)	.34	(.38)
Years of residence	11.24	(11.23)	10.47	(8.48)	4.71	(4.78)	5.04	(4.28)
Central city neighborhood	.31	(.43)	.43	(.42)	.48	(.42)	.48	(.40)
South	.39	(.45)	.31	(.39)	.36	(.40)	.29	(.36)
West	.22	(.38)	.35	(.40)	.34	(.40)	.30	(.36)
Midwest	.23	(.39)	.10	(.26)	.11	(.26)	.13	(.27)
Northeast	.17	(.34)	.24	(.36)	.19	(.33)	.27	(.35)
Time in sample	3.99	(1.85)	4.11	(1.66)	3.82	(1.68)	4.11	(1.62)
Number of person interviews	422,500		33,728		23,533		1,393

Note: The summary statistics were weighted to represent U.S. residents aged 12 and older.

Abbreviation: SD=standard deviation.

Table 2.

Characteristics of household heads by citizenship status, 2017–2018

	U.S.-born citizen		Naturalized citizen		Noncitizen		Refused or don’t know
Characteristics	Mean	(SD)	Mean	(SD)	Mean	(SD)	Mean	(SD)
Latino	.09	(.23)	.36	(.35)	.52	(.37)	.55	(.35)
Asian, non-Latino	.02	(.10)	.30	(.33)	.27	(.32)	.20	(.28)
White, non-Latino	.74	(.36)	.22	(.30)	.14	(.25)	.16	(.26)
Black, non-Latino	.14	(.28)	.10	(.22)	.07	(.18)	.07	(.18)
Other race, non-Latino	.02	(.12)	.02	(.09)	.01	(.07)	.02	(.09)
Age	51.15	(14.64)	52.80	(11.51)	40.32	(9.87)	40.79	(10.12)
Male	.48	(.40)	.51	(.36)	.55	(.37)	.50	(.35)
Married	.46	(.40)	.62	(.35)	.57	(.36)	.54	(.36)
Divorced	.16	(.30)	.12	(.23)	.07	(.19)	.08	(.19)
Separated	.02	(.13)	.04	(.14)	.05	(.16)	.05	(.15)
Widowed	.10	(.24)	.08	(.20)	.03	(.12)	.02	(.10)
Never married	.26	(.35)	.14	(.25)	.29	(.33)	.31	(.33)
Education in years	13.91	(2.16)	13.48	(2.86)	12.45	(3.35)	12.06	(2.71)
Employed	.59	(.40)	.63	(.35)	.68	(.34)	.69	(.33)
Household income	11.61	(3.51)	11.74	(3.03)	10.36	(3.35)	10.49	(3.12)
Household size	2.18	(1.03)	2.65	(1.10)	2.83	(1.19)	2.41	(1.13)
Homeowner	.64	(.39)	.63	(.35)	.30	(.34)	.27	(.31)
Years of residence	12.00	(10.79)	10.62	(7.46)	4.72	(4.37)	4.56	(3.95)
Central city neighborhood	.33	(.38)	.44	(.36)	.50	(.37)	.51	(.35)
South	.39	(.39)	.31	(.33)	.38	(.36)	.33	(.33)
West	.21	(.33)	.33	(.34)	.32	(.34)	.27	(.31)
Midwest	.24	(.34)	.11	(.22)	.12	(.24)	.16	(.26)
Northeast	.16	(.30)	.25	(.31)	.18	(.28)	.24	(.30)
Time in sample	3.25	(1.47)	3.64	(1.40)	2.83	(1.33)	2.86	(1.35)
Number of household interviews	250,404		19,365		12,018		805

Note: The summary statistics were weighted to represent U.S. residential households.

Abbreviation: SD=standard deviation.

Referring first to Table 1, we see that, as expected, Latinos and Asians account for a majority of the naturalized citizen, noncitizen, and ambiguous citizenship status groups, whereas Whites and Blacks account for a majority of the native-born group. The average native-born citizen is 44 years old, which is about four years older than the average noncitizen and those with ambiguous citizenship status, while naturalized citizens have the highest average age of the four groups (about 51 years). U.S.-born citizens are married at a lower rate than the other three groups. And, with regard to socioeconomic standing, U.S.-born citizens and naturalized citizens generally are in better condition, showing higher levels of education, household income, and homeownership than noncitizens and the ambiguous citizenship status group, even though the latter groups are employed at higher rates than the two citizen groups, which is consistent with findings that noncitizens have relatively high rates of employment, though often at lower wages (Maasoumi & Zhu, 2016). Both native-born and naturalized citizens have resided for longer periods of time in their homes than noncitizens and those with ambiguous citizenship status. Native-born citizens are much less likely than the other groups to reside in central cities (31% compared to near 50%). A higher proportion of the naturalized citizen, noncitizen, and ambiguous citizenship status groups live in the South, West, or Northeast than in the Midwest. These patterns reflect population shares by region in the U.S., and they also can be explained by the early locational patterns of immigrants that have influenced the settlement patterns of later arrivals (Rumbaut, 2014). In Table 2, we find very similar results for group comparisons at the household-head level, so the detailed discussion is not repeated here.

Overall, Tables 1 and 2 clearly set apart U.S.-born citizens from those born outside the U.S. and those with ambiguous citizenship status. It is notable that noncitizens and the ambiguous citizenship group have very similar observed characteristics. This suggests that the NCVS respondents who do not respond to the citizenship item or who respond with “don’t know” are more likely to be noncitizens than to be citizens, especially after excluding from this group persons who are likely to be citizens or have a legal residency status, as described in section 2.2.2.

In order to better understand the “refused or don’t know” responses to the NCVS citizenship question, we also examine how this ambiguous status group compares with the characteristics of undocumented immigrants based on analyses conducted by the Migration Policy Institute (MPI) (Gelatt & Zong, 2018; MPI Data Hub, 2020). Because it is very difficult to identify undocumented immigrants directly, researchers have developed several indirect estimation strategies to detect undocumented immigrants in surveys (e.g., Baker, 2017; Fazel-Zarandi et al., 2018; Passel & Cohn, 2018; Van Hook et al., 2015; Warren & Warren, 2013). These include the residual method (subtracting the government-recorded number of legal immigrants from the survey-based number of total foreign-born population to obtain the estimate of the unauthorized population), the survey direct-question method (asking foreign-born survey participants about their immigration status and counting those not indicating legal statuses as undocumented immigrants), and the cross-survey multiple imputation method (pooling data from two nationally representative surveys — the American Community Survey [ACS] and the Survey of Income and Program Participation [SIPP] — to impute the immigration status of noncitizens) (Bachmeier et al., 2014). MPI uses the last of these methods, which has been shown to produce profiles of the unauthorized population that are broadly in line with those published by the Pew Hispanic Center and the Department of Homeland Security, which have applied alternative approaches (Bachmeier et al., 2014). We therefore use the latest MPI estimates (2014–2018) as benchmarks to assess how the NCVS “Refused or don’t know” group compares to undocumented immigrants in the U.S. along several dimensions. The results of the comparison are shown in Table 3. Although the data available from the NCVS and MPI do not align perfectly on a few measures (e.g., unlike the NCVS, several MPI estimates include people younger than 12 years of age, and the MPI data yield information about region of birth but not race and ethnicity), this comparison is instructive.

Table 3.

Comparing persons of unknown citizenship status in the NCVS with undocumented immigrants in another data source

Refused or don’t know NCVS estimates (2017–2018)		Undocumented immigrants Migration Policy Institute (MPI) estimates (2014–2018)
Gender		Gender
Male	51%	Male a	54%
Age (16 years & over)		Age (16 years & over)
16 to 24	11%	16 to 24	16%
25 to 34	26%	25 to 34	32%
35 to 44	27%	35 to 44	30%
45 to 54	18%	45 to 54	14%
55 and over	17%	55 and over	8%
Family structure (ages 15 & over)		Family structure (ages 15 & over)
Married & living with a partner	46%	Married & living with a partner	45%
Reside with no children	66%	Reside with no children	58%
Employment (ages 16 & over)		Employment (ages 16 & over)
Employed	65%	Employed	66%
Homeownership		Homeownership
Homeowner	34%	Homeowner	29%
Education (ages 25 & over)		Education (ages 25 & over)
Less than high school	35%	Less than high school	48%
High school diploma or GED	26%	High school diploma or GED	24%
Some college or associate’s degree	15%	Some college or associate’s degree	13%
Bachelor’s, graduate, or professional degree	24%	Bachelor’s, graduate, or professional degree	19%
Race/ethnicity		Region of birth (Not directly comparable to race/ethnicity)
Latino	58%	Mexico, Central, & South America a	75%
Asian, non-Latino	22%	Asia a	14%
Black, non-Latino	5%	Africa & Caribbean a	6%
White, non-Latino	13%	Europe, Canada, & Oceania a	6%

Note: MPI estimates (MPI Data Hub, 2020) are based on data from the 2014–2018 American Community Survey (ACS) and the 2008 Survey of Income and Program Participation (SIPP).

^aIncluding people younger than 12 years.

Table 3 shows that the distributions for gender, age (16 & over), family structure, employment status, and homeownership status are similar for the NCVS ambiguous citizenship status group and the MPI estimates for undocumented immigrant characteristics. The NCVS ambiguous citizenship status group contains a slightly higher proportion of respondents with Bachelor’s, graduate, or professional degrees and a smaller proportion of those with less than a high school education than in the MPI’s estimate for undocumented immigrants.¹¹ Additionally, the NCVS ambiguous status group appears to contain somewhat fewer persons of Latino ethnic background and a larger share from Asian and European backgrounds.

Overall, while non-respondents to the NCVS citizenship question and those who provide “don’t know” responses likely encompass a mix of people with different statuses, the descriptive data we have presented (c.f., Tables 1–3) suggest that this ambiguous status group exhibits demographic and socioeconomic attributes that are quite similar to the profile of noncitizens, including the characteristics of undocumented noncitizens, which makes them an interesting group for our analyses.¹²

3.2. Difference in Victimization Rates by Citizenship Status

Does citizenship status affect victimization risk, as we have hypothesized? We begin with a bivariate assessment by presenting in Table 4 rates of violence and property victimization for the different groups examined in the study. To help visualize the comparisons, we also represent the estimated rates in Figures 2a and 2b, with 95% confidence intervals shown as well.¹³ For violent crimes, Figure 2a shows that the victimization rate of the foreign born as a whole (12.46 per 1,000) is significantly lower than that of U.S.-born citizens (23.54 per 1,000). Within the foreign-born population, the victimization rate is the lowest among naturalized citizens (10.34 per 1,000), followed by the rate for noncitizens (14.35 per 1,000), with the t-test indicating that the difference is statistically significant at 5% level. Rates for both of these groups are, in turn, significantly lower than that of the ambiguous citizenship (“refused or don’t know”) group (29.15 per 1,000). Interestingly, while the victimization rate of the ambiguous status group (29.15 per 1,000) is also higher than that of the U.S.-born citizens (23.54 per 1,000), the difference is not large enough to make a clear distinction between the two groups, given the large associated confidence interval.

Table 4.

Rates of violent and property victimization by citizenship status, 2017–2018

Type of crime	U.S.-born citizen		Foreign born as a whole (including ambiguous citizenship status)
			All combined		Naturalized citizen		Noncitizen		Refused or don’t know
	Rate a	(SE)	Rate a	(SE)	Rate a	(SE)	Rate a	(SE)	Rate a	(SE)
Violent victimization	23.54	(.94)	12.46	(1.05)	10.34	(1.13)	14.35	(1.56)	29.15	(6.72)
Number of person interviews	422,500		58,654		33,728		23,533		1,393
Property victimization	110.50	(1.70)	88.52	(3.32)	76.82	(3.89)	105.84	(5.72)	117.19	(21.02)
Number of household interviews	250,404		32,188		19,365		12,018		805

Note:

^aAverage yearly rate of victimization per 1,000 residents aged 12 years and older for the 2-year period Abbreviation: SE=standard error.

Figure 2.

Rates of victimization and 95% confidence intervals, by citizenship status, 2017–2018

Note: The horizontal lines represent victimization rates of U.S.-born citizens.

For property crimes, Figure 2b shows a similar overall pattern, though with some notable differences. Consistent with the patterns observed for violence, property victimization rate is lower among the foreign born as a whole than among U.S. born citizens, and naturalized citizens have a lower property victimization rate than do noncitizens. However, deviating from the findings for violence, although the ambiguous status group has a property victimization rate higher than those of naturalized citizens and noncitizens, these group differences are not statistically significant at 5% levels. Overall, the differences observed for property victimization are not as large, compared to those observed for violent crimes. For example, as the horizontal line in Figure 2b indicates, three groups (U.S.-born citizens, noncitizens, and the ambiguous status group) are located close to the line, and the differences between them are small and statistically insignificant. Naturalized citizens are further away from the line, but even for this group, their property victimization rate is only marginally lower (at 10% level) than that of the ambiguous status group.

Taken together, the bivariate results can be viewed as a first confirmation that there are distinct victimization patterns across the groups, and the results indicate that the group differences may be particularly large for violent crimes. To better understand how these differences are related to citizenship status or other demographic and socioeconomic characteristics, we next turn to multivariable analyses that provide test of our hypotheses.

The results for multivariable logistic regression models are presented in Tables 5 (violent victimization) and 6 (property victimization), with all control variables added to the analyses (see Appendix 1 for base models without the control variables). Both tables contain three models, which correspond to the hypotheses H1, H2, and H3, respectively. We find that the control variables exhibit patterns that are highly similar to those reported in several previous studies of victimization risk based on the NCVS (e.g., Lauritsen & Carbone-Lopez, 2011; Park & Fisher, 2017; Rezey, 2020; Xie & Baumer, 2018). For example, as the results for H1 in both Tables 5 and 6 indicate, having higher income, owning a home, being older, being married, living outside of central cities, and living in the Northeast are all related to lower risks of violent and property victimizations. The tables also show that when income and other characteristics are controlled, there are still significant differences in victimization risk across racial-ethnic groups. For example, for both violent and property crimes, Asians are at lower risks of victimization than all other groups (Blacks, Whites, Latinos, other races), whereas the risk is higher for the “other race, non-Latino” group than for any other group. When we limit the analyses only to the relevant samples of foreign-born respondents referenced in H2 and H3, fewer of these control variables show significant effects on victimization risks, as the sample restriction necessarily reduces the variability of the control variables, compared to that of control variables in the full U.S. samples. Still, household income, age, gender, marital status, and one race variable (Asian) continue to significantly influence violent victimization risk (see Table 5), whereas property victimization is affected by age, marital status, household size, central city location, and region (see Table 6).

Table 5.

Logistic regression models of violent victimization, 2017–2018

Characteristics	H1		H2		H3
	b	(SE)	b	(SE)	b	(SE)
Citizenship status
Born U.S. citizen	Reference		--		--
Foreign born as a whole	−.43 ***	(.08)	--		--
Naturalized citizen	--		−.15	(.18)	−.94 *	(.38)
Noncitizen	--		Reference		−.75 *	(.37)
Refused or don’t know	--		--		Reference
Control variables
Household income	−.06 ***	(.00)	−.06 ***	(.02)	−.06 ***	(.02)
Homeowner	−.20 ***	(.05)	.01	(.18)	.04	(.18)
Education in years	−.00	(.01)	.01	(.02)	.01	(.02)
Employed	−.04	(.05)	−.01	(.18)	−.01	(.17)
Age	−.02 ***	(.00)	−.02 **	(.01)	−.02 **	(.01)
Male	.06	(.04)	.35 *	(.14)	.30 *	(.14)
Latino	−.32 ***	(.07)	−.31	(.22)	−.37	(.21)
Asian, non-Latino	−.70 ***	(.13)	−.72 **	(.25)	−.76 **	(.25)
Black, non-Latino	−.25 **	(.07)	−.34	(.33)	−.37	(.32)
Other race, non-Latino	.47 ***	(.10)	−.35	(.54)	−.20	(.47)
Divorced	.86 ***	(.07)	1.04***	(.24)	1.10***	(.23)
Separated	1.12***	(.10)	.98 **	(.28)	.93 **	(.27)
Widowed	.50 **	(.15)	1.27 **	(.42)	1.20 **	(.41)
Never married	.35 ***	(.07)	.10	(.21)	.09	(.20)
Household size	−.02	(.02)	.02	(.05)	.01	(.05)
Years of residence	−.02 ***	(.00)	−.02	(.01)	−.02	(.01)
Central city neighborhood	.20 ***	(.05)	.19	(.16)	.22	(.16)
South	−.01	(.08)	−.01	(.23)	.04	(.23)
Midwest	.14	(.08)	−.60	(.38)	−.47	(.27)
West	.48 ***	(.08)	.26	(.23)	.24	(.23)
Time in sample	−.11 ***	(.01)	−.08 *	(.04)	−.08 *	(.04)
Number of person interviews	481,154		57,261		58,654

Note:

Abbreviation: SE=standard error.

^*p < .05

^**p < .01

^***p < .001 (two-tailed test).

Table 6.

Logistic regression models of property victimization, 2017–2018

Characteristics	H1		H2		H3
	b	(SE)	b	(SE)	b	(SE)
Citizenship status
Born U.S. citizen	Reference		--		--
Foreign born as a whole	−.30 ***	(.05)	--		--
Naturalized citizen	--		−.01	(.07)	−.17	(.27)
Noncitizen	--		Reference		−.15	(.27)
Refused or don’t know	--		--		Reference
Control variables
Household income	−.02 ***	(.00)	.00	(.01)	.00	(.01)
Homeowner	−.11 ***	(.03)	−.07	(.07)	−.05	(.07)
Education in years	.00	(.00)	−.00	(.01)	−.00	(.01)
Employed	−.03	(.03)	.08	(.07)	.09	(.07)
Age	−.01 ***	(.00)	−.02 ***	(.00)	−.02 ***	(.00)
Male	−.02	(.02)	−.01	(.07)	−.02	(.07)
Latino	−.02	(.04)	.20	(.10)	.19	(.10)
Asian, non-Latino	−.30 ***	(.07)	−.19	(.12)	−.19	(.12)
Black, non-Latino	.01	(.03)	−.09	(.16)	−.10	(.16)
Other race, non-Latino	.45 ***	(.06)	.10	(.26)	.07	(.26)
Divorced	.37 ***	(.03)	.32 **	(.03)	.32 **	(.11)
Separated	.40 ***	(.06)	.27 *	(.13)	.28 *	(.14)
Widowed	.23 ***	(.05)	.28	(.16)	.27	(.16)
Never married	.07 *	(.03)	−.01	(.10)	−.02	(.10)
Household size	.18 ***	(.01)	.15 ***	(.02)	.15 ***	(.02)
Years of residence	−.00	(.00)	.01	(.00)	.01	(.00)
Central city neighborhood	.41 ***	(.03)	.22 **	(.07)	.23 **	(.07)
South	.20 ***	(.04)	.17	(.11)	.17	(.11)
Midwest	.17 ***	(.05)	.19	(.14)	.21	(.13)
West	.69 ***	(.04)	.55 ***	(.11)	.55 ***	(.11)
Time in sample	−.16 ***	(.01)	−.19 ***	(.02)	−.19 ***	(.02)
Number of household interviews	282,592		31,383		32,188

Note:

Abbreviation: SE=standard error.

^*p < .05

^**p < .01

^***p < .001 (two-tailed test).

Turning to our hypotheses, Tables 5 and 6 provide support for H1. The coefficients for foreign born as a whole are statistically significant and negative (−.43 for violence, and −.30 for property crime), indicating lower victimization risks compared to U.S. natives. Consistent with arguments in our theory section, the multivariable results for violence suggest that the demographic and socioeconomic control variables partially explain the bivariate differences observed (see Appendix 1a) between foreign-born and U.S.-born respondents. A supplemental analysis of mediation using methods described by Breen, Karlson, and Holm (2013) revealed that differences in homeownership, years of residence, marital status, and exposure to central-city residential neighborhoods were most consequential for this attenuation. For property crime, the bivariate difference observed between foreign-born and native-born respondents (see Appendix 1b) increased after adjusting for group differences in the control variables, which is suggestive of suppression. In this case, applying the Breen et al. (2013) decomposition method yielded evidence that this was largely because of differences in homeownership, years of residence, and central-city location, as well as difference in household size and age of household head. Most pertinent to the arguments presented earlier, the multivariable results reveal that the expected “protective benefits” of being foreign born persist, even with other factors considered, which is consistent with H1.

In the models for H2, we observe no significant difference between naturalized citizens and noncitizens after incorporating the control variables. The coefficients reported for violent victimization (−.15) and property victimization (−.01) are in the expected direction stipulated in H2, but they are not statistically significant. Thus, while the bivariate results (Table 4 and Appendix 1) suggest meaningful differences in victimization risk between naturalized citizens and noncitizens, the multivariable analyses reveal that those differences are largely a function of other attributes reflected in the control variables. Exploring this further with the Breen et al. (2013) approach, we found that differences in household income, age, homeownership, and household size were most consequential for the attenuation of the observed bivariate differences in victimization risk among naturalized citizens and noncitizens. These patterns are consistent with the idea that there is differential selection into naturalized citizenship as shown in different sociodemographic variables (National Academies of Sciences, Engineering, and Medicine, 2015). Most pertinent to our assessment of H2, the hypothesized protective role of being awarded citizenship is small and statistically insignificant for both types of crimes after considering group differences in these control variables.

In the models for H3, the regression analyses confirm the observations made in the bivariate analyses, showing that the ambiguous citizenship status group has a significantly higher risk of victimization by violence, but not by property crimes (which happen at the household level), when compared to naturalized citizens and known noncitizens.¹⁴

To summarize the findings and show the patterns more intuitively, we use two odds ratio plots (Figures 3a and 3b), with 95% confidence intervals, to visualize regression results for violence and property crimes, respectively. In the graphs, the odds ratio of 1 represents U.S.-born citizens, a reference group we use so that the multivariable-analysis results plotted in Figure 3 can be easily compared with the bivariate results as shown in Figure 2.¹⁵ These graphs clearly show that it is important not only to study the foreign-born as a whole group, but also to evaluate variation within the foreign-born population. Naturalized citizens and known noncitizens have similarly lower risk of victimization for both violent and property crimes compared to U.S.-born citizens, after adjusting for control variables. In contrast, there is no significant difference in victimization risk between the U.S.-born citizens and the ambiguous status group in either of the crime types. Parallel findings were observed after excluding non-Latino Whites from the ambiguous status group, which suggests that the observed patterns are not due merely to the possible presence of non-respondents who are White U.S.-born citizens. Nonetheless, it is important to recognize that the confidence intervals for the ambiguous status group are relatively large, so it would be useful to reconsider these comparisons when additional years of data from the NCVS become available.

Figure 3.

Odds ratio of victimization with 95% confidence intervals, controlling for measured confounding variables, 2017–2018

Note: The horizontal lines represent the reference group, U.S.-born citizens.

4. DISCUSSION

By examining the citizenship data collected in the largest victimization survey (NCVS) in the U.S., we find evidence that an immigrant’s foreign-born status, but not their acquired U.S. citizenship, confers a protective effect against criminal victimization (i.e., H1 supported and H2 rejected). We also find that the protective benefit associated with being foreign born does not extend to those with “ambiguous citizenship status,” who experienced higher levels of violent victimization but not property victimization relative to their naturalized citizen and known noncitizen counterparts (i.e., H3 is supported only for violent crimes). These findings, as we show in Figures 3a and 3b, allow us to visualize a continuum of victimization risk, on which naturalized citizens and known noncitizens are placed on the one end of the continuum with a lower risk of victimization, whereas U.S.-born citizens and the ambiguous citizenship status respondents are on the other end of the continuum with a higher risk of victimization.

We acknowledge that immigrants who are in marginalized and vulnerable positions may be reluctant to report certain types of crimes to NCVS interviewers (e.g., Menjívar & Salcido, 2002; Miller, 2008), an issue we discuss in more depth below. Yet, for multiple reasons we have confidence in the patterns observed for our study of overall violence and property crime. The placement of the NCVS citizenship item at the end of the survey instrument should reduce the chances of systematic underreporting related to immigrant and citizenship status. Additionally, our finding of lower overall violent and property crime victimization among the foreign born — including both noncitizens and naturalized citizens — aligns well with a considerable body of research that reveals lower crime rates in communities in which the foreign born compose a larger share of the population (see Ousey and Kubrin, 2018). Further, the patterns showing relatively higher violent victimization rates for the ambiguous citizenship group — a group that is likely the most marginalized and vulnerable — indicate that our results cannot be explained away by underreporting. In an effort to check the threat of response bias, we observed parallel findings when limiting the analyses to crimes for which there are fewer concerns about systematic underreporting in surveys (i.e., limiting the assessment of violence to crimes of robbery and non-sexual forms of assault). As we later discuss, additional research on the measurement of victimization among immigrants is badly needed, but we consider the patterns reported in our study as meaningful evidence of underlying realities in exposure to overall levels of violence and property crime.

Drawing on research on selective migration and theoretical arguments about the impact of immigration on community revitalization, we posited that, all else equal, the foreign born would be less likely to be victimized because of their lower levels of involvement in risky behaviors and greater exposure to “protective” immigrant community contexts. Our findings support this hypothesis, even though the data we used in this study cannot discern the suggested mechanisms. The argument is plausible when integrating our evidence of lower victimization risk among the foreign born with other research findings that show this group is less involved in alcohol/drug use, antisocial conduct, and crime offending behaviors (e.g., Salas-Wright et al., 2018; Vaughn et al., 2014; Bersani & Piquero, 2017) and more likely to reside in communities with a wide range of family, social, and economic capitals as community revitalization theory suggests (Martinez & Lee, 2000; Sampson, 2008). These explanations remain tentative until direct and comprehensive tests of the proposed mechanisms are conducted. Extending the research in this direction should be an important priority for subsequent research.

We found mixed support for our hypotheses about the implications of heterogeneity in the foreign-born population for differences in victimization risk. On the basis of citizenship and social inclusion/exclusion theories (Bloemraad, 2018; Bosniak, 1988; Menjívar & Abrego, 2012), we expected that foreign-born residents may gain more crime protection benefits by becoming U.S. citizens. This expectation was rooted in theoretical and empirical literature that highlights a wide range of benefits for individuals who acquire U.S. citizenship (National Academies of Sciences, Engineering, & Medicine, 2015; Van Hook et al., 2006), but as our multivariable findings show, the anticipated benefits do not appear to extend to a reduction in victimization. We did observe significantly lower rates of victimization among naturalized citizens (compared to noncitizens) in bivariate analyses, similar to findings reported by BJS (Morgan & Oudekerk, 2019; Morgan & Truman, 2018), but our multivariable analyses revealed that those differences were attributable to variation between the two groups in income, age, homeownership, and other socioeconomic factors associated with migrant selection and incorporation experiences in the U.S. (Sumption & Flamm, 2012).

It is possible that citizenship status influences victimization outcomes indirectly through its impact on subsequent socioeconomic status (e.g., by enhancing household income), a possibility we could not test with just two years of NCVS panel data that encompass citizenship information, but all the evidence we present in this study suggests that acquiring U.S. citizenship adds little significant protective benefits against victimization for the foreign born. This is contrary to theoretical expectations, but it is consistent with the findings that victimization risk is highest among the native-born, despite their uniform status as U.S. citizens. The data used for our study do not illuminate why naturalized citizenship fails to yield the expected reductions in victimization risk for the foreign born, but one post-hoc explanation is that increased integration with American culture and associated changes in lifestyle that often accompany naturalization (e.g., increased freedom to pursue leisure or other activities outside the home) may have the unintended consequence of increasing exposure to situations and settings that negate some of the protective functions expected from U.S. citizenship (e.g., better earnings and more legal protection). It would be valuable for future research to explore this and other possible reasons why attaining citizenship status does not appear to translate into lower victimization risk among the foreign born.

Our findings for violence risk among the “ambiguous citizenship status” respondents in the NCVS conform to the theoretical expectations we delineated. We presented evidence that these ambiguous-status respondents have many demographic and socioeconomic features in common with noncitizens, including the characteristics of the undocumented population. While we caution against labeling the ambiguous citizenship group in the NCVS as undocumented immigrants, it is reasonable to assume that this group contains a higher proportion of undocumented immigrants than the known noncitizen group in which all of the respondents are willing (or able) to share their citizenship information with survey interviewers. With that backdrop in mind, it is noteworthy that compared to naturalized citizens and known noncitizens, respondents with ambiguous citizenship status exhibit significantly higher risks of violent victimization, as we hypothesized.

In previous studies, researchers have observed through case studies and field observations that undocumented immigrants are particularly vulnerable to criminal victimization because of limited resources for social mobility and lack of legal protection (e.g., Cleaveland, 2010; Massey, 2007; Menjívar & Abrego, 2012). Many of these studies drew on socioeconomic and lifestyle-routine activity theories to propose that undocumented immigrants may be especially vulnerable to elevated risks of violence or property crime for factors such as physical visibility (e.g., race/ethnicity, insufficient English), potential yield (e.g., use of cash and close substitutes for cash such as prepaid cards and money orders), greater exposure to motivated offenders at work sites or even in homes, lack of guardianship (e.g., unwillingness to call the police), and lack of access to welfare and social services, all of which increase their target suitability for crime (Bucher, Manasse, & Tarasawa, 2010; Fussell, 2011; Grubb & Bouffard, 2014; Zatz & Smith, 2012). Synthesizing these observations with our findings suggests that the ambiguous status group in the NCVS may exhibit higher rates of violent victimization because many of them lack documented legal status, which distinguishes them from their naturalized-citizen and known noncitizen counterparts and increases their vulnerability to crime. The absence of a similar pattern for property crime victimization may occur because, for crimes examined in this study such as burglary, motor vehicle theft, and other types of household theft in which there is no personal contact between the offender and the victim, the victim’s legal documentation status may not be known or be as relevant as it is for violent crimes, which require personal contact.

We cannot rule out the possibility that the higher victimization risks observed for the ambiguous citizenship status group reflects factors besides immigration or documentation status. Although we controlled for many demographic and socioeconomic attributes in the NCVS data, other individual- and structural-level characteristics associated with victimization risk not included in our study may influence a person’s decision to answer the citizenship question. For example, some individuals may avoid answering the citizenship question because they want to protect members of their households or communities who may be harmed by discriminatory behaviors and policies against immigrants (see, e.g.,da Silva Rebelo, Fernández, & Achotegui, 2018; Martinez & Iwama, 2014; Menjívar & Abrego, 2012; Schueths, 2012). Others may avoid the citizenship question because of a more general mistrust of government and a tendency to avoid sensitive topics (Hogan, 2020). Irrespective of legal citizenship status, these factors (e.g., avoidance behaviors and a mistrust of government) may increase one’s vulnerability to victimization (Kirk & Matsuda, 2011; Ruback & Menard, 2001; Sampson & Bartusch, 1998), which may create or reinforce differences in victimization risks observed in our study.

Deciphering with certainty the factors that may account for the higher rates of violent victimization observed among those with ambiguous citizenship status in the NCVS will be challenging, but this is an important avenue for future research. Additional insights may be gained by extending the present study to consider the potential moderating role of individual-level factors such as racial-ethnic identity and socioeconomic status, which may be feasible when additional years of data become available. Beyond this, it would be valuable to incorporate data on characteristics of the communities in which the respondents reside (e.g., census tracts, cities, counties). The public-use NCVS data used for the present study do not contain geographic data, but this additional information has on occasion been made available to researchers (e.g., Lauritsen, 2001; Lauritsen & Carbone-Lopez, 2011; Xie & Baumer, 2018). Extending such access to the NCVS data that encompasses citizenship would permit many useful extensions of our analysis. For example, by integrating information about the estimated share of the population in respondents’ counties who are undocumented immigrants, the research could supply additional context about whether the ambiguous citizenship status respondents in the NCVS are likely to be undocumented immigrants based on where they live. Further, incorporating additional contextual information at different geographic levels (e.g., state, county, census tract), including indicators of socioeconomic deprivation, intensive immigration enforcement, and other conditions that may promote legal cynicism (Anderson, 1999; Kirk & Matsuda, 2011; Sampson & Bartusch, 1998; Zatz & Smith, 2012), may reveal insights about the underlying structural mechanisms by which NCVS respondents with ambiguous citizenship status experience higher victimization rates. The BJS has historically archived restricted-use NCVS data within the U.S. Federal Statistical Research Data Centers that permit this type of multilevel analyses of victimization (Lauritsen, 2001; Xie & Baumer, 2018), and we hope that they will do so in the future for data that contain respondent citizenship status.

We also encourage future research that assesses the relationship between immigrant/citizenship status and specific crimes of violence, such as rape, sexual assault, and domestic violence, which could reveal patterns that diverge from the evidence reported in our paper. These crimes comprised a very small portion of the crimes captured in our victimization measures, and the limited number of years of available data did not provide sufficient statistical power to replicate our analysis separately for these very serious incidents. Extending the analysis in this direction may be feasible as more data become available, though we also see a need for further refinements to the NCVS to support rigorous investigations of crimes long known to be susceptible to underreporting in surveys (Fisher & Cullen, 2000; Miller, 2008; National Research Council, 2014). The NCVS was redesigned in the early 1990s in an effort to more accurately measure rape, sexual assault, and crimes perpetrated by persons known to respondents, and there is evidence that such changes uncovered many incidents that were previously undetected (Kindermann, Lynch, & Cantor, 1997). Nonetheless, a comprehensive review suggested that the NCVS could be further improved by adopting behaviorally specific questions or using self-administered procedures (National Research Council, 2014). Integrating such survey strategies into the NCVS could prove valuable for further enhancing the validity of crimes that immigrants and others in vulnerable positions, especially women, may be otherwise reluctant to report to interviewers (Menjívar & Salcido, 2002; Miller, 2008). More generally, we strongly encourage future victimization data collection efforts, including the NCVS, to engage with issues of immigration, legal status, gender, race/ethnicity, class, and social norms of crime reporting in their efforts to develop more contextually and culturally sensitive measures of crime and victimization.

In closing, the findings presented in this study underscore the importance of exploring how patterns of victimization vary between U.S.-born citizens and different groups of immigrants. Increasing knowledge about such differences is critical for informing policy and practice in current days, a need that has been stressed even more as we are challenged to think about immigration policies in the context of heightened opposition toward immigration and the past and continuing efforts by some government officials to control immigration, often justified by unfounded claims that immigrants are a major part of the U.S. crime problem (Miroff, Sacchetti, Hauslohner, & Hernández, 2020). Our study adds new evidence to discussions about this important issue. By showing that the majority of immigrants (i.e., naturalized citizens and known noncitizens) are protected from victimization by their foreign-born status, we join a long line of research that challenges claims that crime is more prevalent among immigrants and within immigrant communities (e.g., Bersani & Piquero, 2017; Butcher & Piehl, 1998; Desmond & Kubrin, 2009; Eggers & Jennings, 2014; MacDonald, Hipp, & Gill, 2013; Martinez & Lee, 2000; Miller & Peguero, 2018; Ousey & Kubrin, 2018; Xie & Baumer, 2018; Zatz & Smith, 2012). This body of accumulated knowledge, coupled with empirical evidence showing the ineffectiveness of the government’s main immigration control measures such as 287(g), Secure Communities, border security, and the detention and deportation of immigrants (e.g., Forrester & Nowrasteh, 2018; Kubrin, 2014; Miles & Cox, 2014; Treyger, Chalfin, & Loeffler, 2014), suggests that the intent to use immigration control as a method for reducing crime is likely to be impractical and may bring unintended consequences. Now with the global health and economic crisis, the need for keeping people safe and using resources more efficiently is more urgent than ever. We need to understand, for example, the mechanisms that protect the foreign born but not U.S.-born citizens so we can learn from one another and be better informed on crime control strategies. Also, by exploring the factors that make some groups of immigrants such as the group with ambiguous citizenship status to experience considerably higher risk of violent victimization, we may begin to identify the individual or structural factors contributing to the life conditions of these individuals and build a safe environment for all.

Appendix 1. Logistic regression models of violent and property victimization, 2017–2018, base models

A1a.

Violent victimization, base model (without controls)

Characteristics	H1		H2		H3
	b	(SE)	b	(SE)	b	(SE)
Citizenship status
Born U.S. citizen	Reference		--		--
Foreign born as a whole	−.63 ***	(.07)	--		--
Naturalized citizen	--		−.51 **	(.14)	−1.01***	(.25)
Noncitizen	--		Reference		−.51 *	(.25)
Refused or don’t know	--		--		Reference
Number of person interviews	481,154		57,261		58,654

Note:

Abbreviation: SE=standard error.

^*p < .05

^**p < .01

^***p < .001 (two-tailed test).

A1b.

Property victimization, base model (without controls)

Characteristics	H1		H2		H3
	b	(SE)	b	(SE)	b	(SE)
Citizenship status
Born U.S. citizen	Reference		--		--
Foreign born as a whole	−.17 ***	(.04)	--		--
Naturalized citizen	--		−.37 ***	(.06)	−.07	(.21)
Noncitizen	--		Reference		−.05	(.22)
Refused or don’t know	--		--		Reference
Number of household interviews	282,592		31,383		32,188

Note:

Abbreviation: SE=standard error.

^*p < .05

^**p < .01

^***p < .001 (two-tailed test).