Surveillance and system avoidance

Drawing on scholarship on surveillance (Foucault 2007 [Reference Foucault2007 [1977]]), procedural justice (Tyler Reference Tyler2006), and legal cynicism (Sampson and Jeglum Bartusch Reference Sampson and Jeglum Bartusch1998), system avoidance theorizes that individuals concerned about formal social control by the state or its authorities are likely to evade surveilling institutions that keep formal records of individuals’ transactions with them (Brayne Reference Brayne2014; Goffman Reference Goffman2009). This research locates concerns about formal social control in a person’s sanctionable status, an individual-level characteristic that signals either a past or potential negative interaction with state authorities. The sanction may reflect the enforcement of civil or criminal law, particularly if the boundaries between them are blurry (e.g., Beckett and Herbert Reference Beckett and Herbert2009; Stumpf Reference Stumpf2006, see also Young et al. Reference Young, Martin and Lageson2024).Footnote ² Past criminal-legal contact is the most typical sanctionable status considered (Brayne Reference Brayne2014; Haskins and Jacobsen Reference Haskins and Jacobsen2017; Lanuza and Turney Reference Lanuza and Turney2020). Evidence of system avoidance is also found among individuals vulnerable to deportation (Desai et al. Reference Desai, Jessica Houston and Adelman2020; Dondero and Altman Reference Dondero and Altman2022; Patler and Gonzalez Reference Patler and Gonzalez2020; Wong et al. Reference Wong, Shklyan, Isorena and Peng2019), a sanctionable status thought to be similar to criminal-legal contact (e.g., Brayne Reference Brayne2014: p. 387; Goffman Reference Goffman2009: p. 354, see also Waters and Kasinitz Reference Waters, Kasinitz, Foner and Simon2015).

One reason system avoidance may occur is because individuals with a sanctionable status mistrust government and worry that law enforcement officers in their local context may seek to apprehend them. Surveilling institutions – namely, those concerned with medical care, finances, employment, and education – help facilitate this apprehension because they are legally required to catalog individuals’ interactions with them; they may, therefore, offer law enforcement a paper or digital trail to apprehend wanted individuals (Brayne Reference Brayne2020). System avoidance owing to the threat of formal social control is theorized to occur when individuals vulnerable to sanction have lower average rates of involvement in surveilling institutions relative to individuals who are not vulnerable. That individuals vulnerable to sanction have similar rates of involvement in non-surveilling (i.e., religious or volunteer) institutions, relative to individuals who are not vulnerable, offers additional evidence linking a sanctionable status to system avoidance (Brayne Reference Brayne2014).

Although a sanctionable status signals vulnerability to formal social control, its meanings for the individuals holding it vary. Ethnographic research notes that, while a sanctionable status may sometimes amount to a “master status” that dominates individuals’ routines – including their relationships with surveilling institutions – they are seldom totalizing (c.f. Hughes Reference Hughes1945). On the one hand, it may be that individuals sharing a sanctionable status report such extreme fears of formal social control that they will avoid any institution where the police might find them, as Goffman (Reference Goffman2009) outlines in her study of young Black men with arrest warrants in a Philadelphia neighborhood. Yet Goffman’s respondents maintained some involvement despite being wanted by the law. For Goffman, a sanctionable status may offer individuals a reason to avoid fulfilling responsibilities associated with their social roles (e.g., as parents). On the other hand, it may also be that the meanings of a sanctionable status are more complex than initially suggested: a sanctionable status may imply some system noninvolvement, but it may also imply new or maintained system involvement. Rios (Reference Rios2011) observes that young men of color who have criminal-legal contact in a poor Oakland neighborhood resist having been policed by becoming civically engaged. Recent ethnographies likewise suggest that involvement and noninvolvement are situational among individuals with a sanctionable status (e.g., Bell Reference Bell2016; Reference Bell2020; Stuart Reference Stuart2016; for reviews, see Rios et al. Reference Rios, Carney and Kelekay2017; Stuart et al. Reference Stuart, Armenta and Osborne2015).

Similar dynamics of system involvement alongside noninvolvement emerge in studies of individuals vulnerable to deportation. The threat of deportation can certainly compel some undocumented immigrants to withdraw from daily life, leaving their homes only for absolute necessities in order to minimize the threat of formal social control (Schmalzbauer Reference Schmalzbauer2014). Yet, given that 62% of undocumented immigrants have lived in the United States for over 10 years (Migration Policy Institute 2019), many see the threat of deportation as endemic to daily life (Chauvin and Garcés‐Mascareñas Reference Chauvin and Garcés‐Mascareñas2014; De Genova Reference De Genova2002).Footnote ³ Undocumented immigrants not only venture out of their homes but also report involvement in institutions that may surveil them (see Asad Reference Asad2023). They must “learn to be illegal” (Gonzales Reference Gonzales2011), adjusting their routines within the confines of their legal status. Sometimes this means involvement in surveilling institutions, such as when receiving emergency medical care (Van Natta Reference Van Natta2019; Reference Van Natta2023a; Reference Van Natta2023b), paying income tax (Asad Reference Asad2023), working in formal and informal labor markets (Armenta Reference Armenta2024; Rosales Reference Rosales2020; Somashekhar Reference Somashekhar2022), and/or enrolling in or participating in their kids’ schooling (Bruhn Reference Bruhn2023; Cuevas Reference Cuevas2021; Enriquez Reference Enriquez2017; Pedroza and Losada Reference Pedroza and Losada2025; Ruszczyk Reference Ruszczyk2021).

Statistical analyses of survey data have evaluated several possible explanations for the involvement and noninvolvement uncovered in this ethnographic literature. Perhaps the most common is the multiple social roles that individuals with sanctionable statuses hold. Although Brayne (Reference Brayne2014) shows system avoidance to be a plausible mechanism underlying lower average rates of involvement in surveilling institutions among individuals with, relative to without, criminal-legal contact, her test excludes mothers with young children in analyses of involvement in educational and/or employment institutions (because “having a child under 12 years of age … is the strongest predictor, for women, of being neither in school nor working,” p. 374). Parenthood is nonetheless an important factor predicting involvement – especially for mothers – in educational institutions on behalf of one’s children among individuals with and without incarceration experience (Haskins and Jacobsen Reference Haskins and Jacobsen2017). And similar dynamics are observed among individuals vulnerable to deportation, with Latino noncitizens showing lower average rates of involvement in surveilling institutions relative to the Latino U.S.-born population on their own behalf – but similar rates on behalf of their children (Asad Reference Asad, van Hook and King2024; Dondero and Altman Reference Dondero and Altman2022).

One less considered explanation is that rates of institutional involvement in part reflect spatial and temporal variation in the threat of formal social control, as measured by the laws, policies, or practices that facilitate state sanction. Although ethnographic studies have noted as much, statistical tests do not explicitly consider the threat of formal social control, rely on data from a single spatial and/or temporal context, or examine only a single surveilling institution. For example, Remster and Kramer (Reference Remster and Kramer2018) offer the first statistical test of whether spatial and temporal contexts mediate the relationship between a sanctionable status and system involvement. They replicate Brayne’s (Reference Brayne2014) analysis and add characteristics that proxy the structural conditions of survey participants’ neighborhoods (e.g., racial composition). Though they find that, net of these characteristics, system avoidance remains more likely among individuals with criminal-legal contact, relative to those without, they do not explicitly measure the threat of formal social control.Footnote ⁴ When scholars do consider spatial and temporal variation in the threat of formal social control, they tend to study one surveilling institution – usually medical or educational (e.g., Dee and Murphy Reference Dee and Murphy2018; Dondero and Altman Reference Dondero and Altman2022) – and show inconsistent evidence of noninvolvement associated with the threat of formal social control.Footnote ⁵ But, following Brayne’s (Reference Brayne2014) theoretical formulation of system avoidance, consideration of a range of surveilling institutions is important because noninvolvement in one does not necessarily foretell noninvolvement in another (Asad Reference Asad2023). The few studies that consider multiple surveilling institutions represent a single spatial and temporal context, as in Desai et al.’s (Reference Desai, Jessica Houston and Adelman2020) study of Los Angeles in 2004 and Wong et al.’s (Reference Wong, Shklyan, Isorena and Peng2019) study of San Diego in 2017. While these are necessary contributions, the literature is missing an analysis of how individuals sharing a sanctionable status relate to a range of surveilling institutions across spatial and temporal contexts of formal social control.

Spatial and temporal contexts of policing

Although a sanctionable status is one indicator of an individual’s vulnerability, the threat of formal social control for individuals sharing a sanctionable status varies across space and time. Specifically, we argue that this threat varies in part by the laws, policies, and practices that magnify or constrain policing in daily life. Unlike durable structural factors such as racial/ethnic composition or concentrated disadvantage, the laws, policies, and practices governing policing “can be reorganized with the stroke of a pen, [their] philosophies…can be uprooted over fairly short periods of time, [and their] practices … can be trained away or deterred” (Bell Reference Bell2020: 924). Spatial and temporal dynamism in the laws, policies, and practices that threaten to police individuals sharing a sanctionable status may, in turn, reflect broader sociolegal orientations toward formal social control (Edwards Reference Edwards2016). If so, these orientations may help explain whether and to what degree individuals sharing a sanctionable status engage with surveilling institutions.

Policing follows a spatial and temporal pattern. This pattern disproportionately burdens race- and class-subjugated populations throughout the United States – namely, poor and working-class Black and Latino people (Abrego et al. Reference Abrego, Coleman, Martínez, Menjívar and Slack2017; Armenta Reference Armenta2017; Bell Reference Bell2017; Ray et al. Reference Ray, Powelson, Fuentes and Doan2024; Weaver and Lerman Reference Weaver and Lerman2010). The theoretical sources of this burdening vary (see Simes et al. Reference Simes, Beck and Eason2023: pp. 224-226 for an extended review). Some scholars emphasize that policing concentrates in these spaces because that is where crime rates are highest; most police departments in the country use this “hot spots” policing strategy (Braga et al. Reference Braga, Papachristos and Hureau2014). Others clarify that crime rates reflect the policing of overlapping forms of material and social disadvantage; in this perspective, crime rates overlap with rates of poverty, unemployment, and public assistance receipt (Travis, Western and Redburn Reference Travis, Western and Stevens Redburn2014). And still others theorize policing as a mechanism for the surveillance and criminalization of race- and class-subjugated individuals, irrespective of crime rates (Bell Reference Bell2017; Miller and Stuart Reference Miller and Stuart2017); in this perspective, attention is turned to policing strategies, criminal codes, public ordinances, and sentencing practices rather than crime rates per se (Simes et al. Reference Simes, Beck and Eason2023: p. 226). We accord with the third perspective in this article but account for the other two theoretical perspectives throughout our analysis. All theoretical perspectives nonetheless agree that policing’s spatial concentration targets race- and class-subjugated individuals. Over time, criminal-legal contact – including police stops and arrests that may ultimately lead to conviction and incarceration – may come to represent shared experiences for large swaths of race- and class-subjugated populations (Clair Reference Clair2020; Enns et al. Reference Enns, Youngmin, Comfort, Goldman, Lee, Muller, Wakefield, Wang and Wildeman2019; Muller and Roehrkasse Reference Muller and Roehrkasse2022; see Ray et al. Reference Ray, Powelson, Fuentes and Doan2024 for a review), even if trends in some of these outcomes (e.g., racial disparities in incarceration) have begun to reverse in recent data (Robey et al. Reference Robey, Massoglia and Light2023).

As an instantiation of policing more generally, immigration policing also follows a spatial and temporal pattern (e.g., Hausman Reference Hausman2022; Moinester Reference Moinester2018; Varsanyi et al. Reference Varsanyi, Lewis, Marie Provine and Decker2012). It disproportionately burdens poor and working-class Latino noncitizens, especially the undocumented, who have settled in the United States (Golash-Boza Reference Golash-Boza2015). It also burdens Latino citizens, whether naturalized or U.S. born, due to their family relationships and/or being misrecognized as undocumented (Asad and Clair Reference Asad and Clair2018; Flores and Schachter Reference Flores and Schachter2018). The theoretical sources of this burdening are similar to policing more generally, though scholars also emphasize that immigration policing reflects the growing consolidation of the immigration and criminal-legal systems (Asad and Clair Reference Asad and Clair2018; Jain Reference Jain2019; Menjívar Reference Menjívar2021; Stumpf Reference Stumpf2006). This is because, since 1996, the federal government has enlisted state and local police as collaborators in immigration enforcement. Through the Illegal Immigration Reform and Immigrant Responsibility Act’s 287(g) agreements, for example, police officers deputized as immigration officers can deport immigrants in police custody in dozens of counties in 19 states (Menjívar and Kanstroom Reference Menjívar and Kanstroom2013).Footnote ⁶ More pervasive collaborations – namely, the Criminal Alien Program, Secure Communities, and the Priority Enforcement Program – institutionalized police participation in immigration enforcement nationwide between 2008 and 2013 (Golash-Boza Reference Golash-Boza2015), and in the years thereafter. These programs work through database linkages, whereby a police officer fingerprints someone they arrest and runs those fingerprints through law enforcement databases. This search may result in a “detainer” request, a notice that the Department of Homeland Security (DHS, which houses Immigration and Customs Enforcement [ICE]) issues to police to hold someone for an additional 48 hours beyond their scheduled release on the suspicion that they are deportable. Police need not comply with all ICE detainer requests, especially in jurisdictions with a sanctuary policy that circumscribes at least some police cooperation with ICE (Hausman Reference Hausman2020). A detainer nonetheless means that the police have arrested someone whom ICE suspects is deportable. Arrests from state and local police account for most deportations occurring from inside (rather than at the borders of) the country (Capps et al. Reference Capps, Chishti, Gelatt, Bolter and Ruiz Soto2018; Hausman Reference Hausman2020). We term this “immigration policing” but reiterate that it reflects that police officers have arrested someone whom ICE may wish to deport.

Legal consciousness and system involvement

Given spatial and temporal variation in the threat of policing generally, and immigration policing specifically, law and society scholars have drawn on theories of legal consciousness to understand how individuals worried about policing perceive and respond to it in daily life (e.g., Andrews Reference Andrews2018; Bell Reference Bell2016; Flores et al. Reference Flores, Escudero and Burciaga2019; Rocha Beardall Reference Rocha Beardall2024; Stuart Reference Stuart2016). Legal consciousness describes how individuals understand, experience, and respond to legal authorities, institutions, and processes (Chua and Engel Reference Chua and Engel2019; Ewick and Silbey Reference Ewick and Silbey1998; Merry Reference Merry1990; Silbey Reference Silbey2005). Legal consciousness does not just refer to idiosyncratic meanings about the law; it also considers how institutional power across space and time is linked to individuals’ variable orientations toward the law (Silbey Reference Silbey2005). Despite their shared interest in law’s power and meanings in daily life, theories of legal consciousness and system avoidance have developed largely in parallel (but see Asad Reference Asad2023). We argue that legal consciousness offers a productive framework for clarifying whether and how the threat of formal social control matters for institutional involvement among individuals sharing a sanctionable status.

Theories of legal consciousness outline the multiple orientations individuals may hold toward the law, depending on situational context: “against the law” or “with the law” (Ewick and Silbey Reference Ewick and Silbey1998: pp. 39–40). In our adaptation, for individuals exhibiting an “against the law” orientation, prior interactions or experiences with legal authorities, institutions, and processes suggest that the law is dangerous and should be avoided. This orientation aligns closely with current theorization on system avoidance, indicating that individuals sharing a sanctionable status are likely to have lower average rates of involvement in surveilling institutions compared to those who do not hold that status. In the context of both policing and immigration policing, this orientation implies that individuals with criminal-legal contact and/or deportation experience may be particularly inclined to avoid surveilling institutions relative to those without such contact or experience. But many individuals in the United States do not have criminal-legal contact or deportation experience. Members of racialized groups who are disproportionately policed or deported may nonetheless perceive their spatial and temporal context as threatening. This perception can arise from laws, policies, and practices that magnify the threat of policing for individuals sharing their sanctionable status, as suggested by the literature on system avoidance. If so, we might expect lower average rates of institutional involvement where and when the threat of formal social control for individuals holding a given sanctionable status is higher rather than lower. Ethnographic research is agnostic about which institutions might show the lowest involvement, as individuals may “cultivate unpredictability” by limiting involvement in multiple institutions (Goffman Reference Goffman2009: p. 340).

By contrast, individuals may exhibit a “with the law orientation.” While they recognize the potential threats posed by legal authorities, institutions, and processes, those sharing a sanctionable status also understand that these entities allow them “a set of permissible roles, transactions, and purposes” (Ewick and Silbey Reference Ewick and Silbey1998: p. 136). This orientation follows ethnographic studies of race, crime, and justice, which reveal that individuals concerned about policing may still turn to law enforcement for assistance – including help navigating surveilling institutions (Bell Reference Bell2016; for reviews, see Rios et al. Reference Rios, Carney and Kelekay2017; Stuart et al. Reference Stuart, Armenta and Osborne2015). It also resonates with multimethod research highlighting the situational tensions of individuals with criminal-legal contact or deportation experience. As the ethnographic research reviewed in the first section of this literature review suggests, some may forgo routine medical care but access emergency medical care; others may forgo the formal labor market but find ways to work; and others still may forgo formal schooling but enroll or otherwise support their children’s educational pursuits. This may occur because they accept the risks of institutional involvement, recognize the value of continued involvement, or seek to challenge negative stereotypes through their involvement (e.g., Andrews Reference Andrews2018; Armenta Reference Armenta2024; Armenta and Sarabia Reference Armenta and Sarabia2020; Asad Reference Asad2023; Bell Reference Bell2016; Fong Reference Fong2019; Hughes Reference Hughes2019; Rios Reference Rios2011; Stuart Reference Stuart2016). Regarding policing and immigration policing, a with the law orientation may suggest that some individuals sharing a sanctionable status will maintain their involvement in surveilling institutions, regardless of the prevailing threat of formal social control.

Whether a person holds an against or with the law orientation, there may be other laws, policies, and practices that constrain the threat of formal social control. In the context of immigration policing, sanctuary policies, which limit at least some police collaboration with ICE, are one example. Although there are many types of sanctuary policies (Ortiz et al. Reference Ortiz, Farrell-Bryan, Gutierrez, Boen, Tam, Yun, Venkataramani and Montoya-Williams2021), of interest here are those that circumscribe the conditions under which police may cooperate with detainer requests. Sanctuary policies seldom prohibit all police cooperation with ICE; instead, they prioritize “serious” crimes and individuals with repeat offenses (Ascherio Reference Ascherio2022; Hausman Reference Hausman2020). In so doing, they can help eliminate barriers to help-seeking, as research based on Latinos’ crime-reporting behaviors suggests (Martínez-Schuldt and Martínez Reference Martínez-Schuldt and Martínez2021). Yet, in not prohibiting all police collaboration with ICE, Latinos across nativity and citizenship status groupings may still be concerned about immigration policing – even in jurisdictions with sanctuary policies (Houston Reference Houston2019; Lasch et al. Reference Lasch, Chan, Eagly, Francesca Haynes, Lai, McCormick and Stumpf2018). We nonetheless expect that the threat of immigration policing is higher where and when sanctuary policies are absent rather than present, such that reductions in the incidence of involvement in surveilling institutions should be more pronounced in non-sanctuary jurisdictions. And, as above, the conditional degree of involvement should not vary.

Other laws, policies, and regulations may exist independent of the threat of formal social control. If so, individuals sharing a sanctionable status may not “forgo” involvement in surveilling institutions but rather be denied access to them as a matter of law. For example, some states prohibit individuals with certain types of criminal-legal contact, such as convictions related to fraud, from holding a bank account (Brayne Reference Brayne2014: 384). Likewise, some states bar many individuals vulnerable to deportation – especially but not exclusively the undocumented (Watson and Thompson Reference Watson and Thompson2022) – from accessing public health insurance except in emergent circumstances (Van Natta Reference Van Natta2023b), receiving financial aid for college tuition (Cuevas Reference Cuevas2021; Gonzales Reference Gonzales2015), and working without added federal scrutiny of their work authorization (Amuedo-Dorantes and Bansak Reference Amuedo-Dorantes and Bansak2012). These other laws, policies, and regulations create situations in which someone may be denied from accessing an institution on their own behalf but not on behalf of a family member (Asad Reference Asad2023; Dondero and Altman Reference Dondero and Altman2022). They, therefore, may have their own relationships with whether and to what degree individuals with a sanctionable status become involved in surveilling institutions. Regarding the incidence of involvement, we expect lower rates of involvement in surveilling institutions when: (1) the threat of formal social control is higher relative to lower and (2) access to surveilling institutions for individuals sharing a sanctionable status is restricted relative to not. Regarding the conditional degree of involvement, we expect some individuals to maintain their involvement, regardless of the prevailing threat of formal social control or restrictions on their own institutional access.

Taken together, the relationship between a sanctionable status and system involvement may vary situationally according to the threat of formal social control across space and time. Specifically, and drawing on theories of legal consciousness, we posit that spatial and temporal differences in the laws, policies, and practices that threaten to police individuals who share a sanctionable status are associated with the incidence and conditional degree of these individuals’ involvement in surveilling institutions. Using the case of immigration policing and its racialized effects on Latinos across nativity and citizenship status groupings, we ask: Does spatial and temporal variation in the threat of immigration policing matter for the incidence and conditional degree of Latinos’ institutional involvement? If so, does this association vary by nativity, citizenship status, or institutional type? Finally, are there other laws, policies, and regulations that govern access to surveilling institutions that moderate the relationship between immigration policing and the incidence and conditional degree of Latinos’ reported institutional involvement?

Data

We rely on the ATUS, compiled by the IPUMS Center for Data Integration and the Maryland Population Research Center (Hofferth et al. Reference Hofferth, Flood, Sobek and Backman2020), to measure involvement in surveilling institutions among Latinos overall and by nativity and citizenship status across the United States between 2008 and 2019. Fielded each year by the U.S. Census Bureau, the ATUS is a time-diary study sponsored by the Bureau of Labor Statistics. Respondents are selected using a three-stage stratification process from a random subset of households that finished participating in the Current Population Survey (CPS) 2–5 months prior.Footnote ⁷ First, because the CPS oversamples less-populous states, the ATUS removes some households from these states to improve the efficiency of national estimates. Second, the remaining households are stratified by the race/ethnicity of the householder; the presence and age of any children in the household; and the number of adults in adult-only households. Households with Latino householders, and households with children, are oversampled. Finally, within each sampled household, one member aged 15 or older is randomly selected to complete a 1-day time diary.

The time diaries document where and how ATUS respondents spend their days. They cover each day of the week, and weekends are oversampled. In each household, a single respondent catalogs all activities from 4:00 AM the previous day to 4:00 AM on the interview day for an interviewer, in English or Spanish. The interviewer records the activities verbatim. All activities are later classified into 17 major categories (e.g., education, household activities, traveling) that span 400 subcategories.

Time diaries afford several advantages to a study of the relationship between the threat of formal social control and system involvement across space and time among individuals sharing a sanctionable status. In contrast to measures that ask for recall of events farther into the past, time diaries more reliably capture recent behavior – allowing us to examine what individuals do in daily life (Meyer et al. Reference Meyer, Mok and Sullivan2009). The time diaries also capture a range of activities related to respondents’ involvement in surveilling institutions. For instance, having a checking account is one indicator of institutional involvement typical in research on system avoidance. It is nonetheless possible for someone to lack a checking account and to visit a bank and consult with a banker or teller, among other activities. The time diaries include this range in our estimates. Still, we cannot account for situations where, for example, respondents have a checking account but did not report any involvement in financial institutions in the last 24 hours.

The ATUS is a valuable dataset for studying the threat of formal social control and system involvement across space and time, but there are limitations to the sampling design worth noting. One is that, though the ATUS samples households at the state level and oversamples households of interest to this study, eight states in the data have very few Latino respondents (and sometimes have no Latino respondents in a survey year).Footnote ⁸ This means that, though our data include respondents across 42 states, these tend to be states with larger populations and more Latinos. In addition, since time diaries are collected from a single, randomly-selected household member for a single, 24-hour period, they do not allow us to examine within-household or -person changes over time. Finally, as a survey of the civilian population, the ATUS does not include any individuals whom law enforcement agencies have detained by the start of the survey.

Although we apply sampling weights provided by the ATUS to correct for survey non-response, oversampling, and design-specific issues (e.g., the day of the survey), these other characteristics of the data may influence our estimates in several ways. The inclusion of states with higher shares of Latinos may lead us to overestimate institutional involvement because places with smaller Latino populations are missed. In these so-called new destinations, to be Latino often means to be undocumented, rendering institutional involvement more fraught (Schmalzbauer Reference Schmalzbauer2014). Yet, our results may also underestimate institutional involvement for the same reason, as qualitative research shows that many new destinations promote new arrivals’ institutional involvement through both formal policies and informal practices (Marrow Reference Marrow2011; Williamson Reference Williamson2018). We further suspect that our estimates will be off if the ATUS misses individuals who were detained prior to their household’s selection for the survey. The consequences of this limitation are unknown. On the one hand, we will overestimate involvement if those individuals lived alone and had minimized their institutional involvement prior to their arrest (though, as McConnell Reference McConnell2015 notes, such solitary residence is uncommon). On the other hand, we will underestimate involvement if those arrested individuals in the ATUS sampling frame were actively involved in one or more surveilling institutions. Everyone completing a time diary nonetheless survived the 24-hour period preceding the survey. Our analysis cannot identify a causal relationship, but we draw inferences from ATUS respondents in states with a sizable population of Latinos since 2008 to offer a robust account of the relationship between immigration policing and institutional involvement.

We merge data on Latino respondents’ institutional involvement from the ATUS with state-year data on immigration policing. In our main analyses, these data come from the TRAC at Syracuse University (Long and Burnham Reference Long and Burnham2024). These data provide detailed information from ICE on all detainer requests that the agency issues to state and local police. As explained above, detainers are a measure of state and local police arrests of noncitizens whom ICE has expressed an interest in deporting (Armenta Reference Armenta2017). Although these data do not contain information on direct efforts by federal immigration officers – whether from Customs and Border Protection (CBP) or ICE – to apprehend noncitizens in the United States, they reflect the everyday forms of policing encountered by a settled population of Latinos that, in turn, may lead to their being turned over to ICE (Asad Reference Asad2023; Legewie et al. Reference Legewie, Hsin, Harder and Martén2022). We refer to this as immigration policing. If CBP and ICE engage in other efforts to apprehend noncitizens inside the country, we would underestimate any identified relationship between immigration policing and institutional involvement.

Our time period of interest is between 2007 and 2018, a range of dates that precedes the years included in the ATUS data by 1 year to minimize the risk of reverse causality. During this time, detainers proliferated across Republican and Democratic administrations alike. Our data include 2,277,172 detainer records, which span 2,513 counties across all 50 states and 4 territories (Guam, Northern Mariana Islands, Puerto Rico, and the U.S. Virgin Islands). The TRAC data are missing geographic information for 6.53% of all detainer records (n = 148,781), ranging from a low of 1.38% in 2018 to a high of 16.86% in 2008.

Finally, we rely on additional data sources to measure other aspects of state contexts that may bear on any observed relationship between immigration policing and institutional involvement. We explain these measures below and in Appendix Table A1. For control variables, we draw on data from the American Community Survey; the Annual Survey of State and Local Government Finances compiled by the Urban Institute; the U.S. Census; the Vera Institute of Justice’s Incarceration Trends dataset; and the Federal Bureau of Investigation’s Uniform Crime Reporting Program Data as compiled by Kaplan (Reference Kaplan2021). We also consider state-year data on sanctuary policies and, given the results of the main analysis, policies that govern noncitizens’ access to the labor market and educational institutions.

In linking the ATUS with data from TRAC and other sources, we make two assumptions. First, we assume that the state in which ATUS respondents are surveyed reflects the context within which any reported institutional involvement occurred. Second, we assume our estimates represent a lower bound for the variance of spatial effects on institutional involvement. There is variation within states (i.e., at the county or municipal levels) across measures used in this analysis, particularly those related to immigration policing. The state nonetheless represents the ATUS’ lowest aggregation of population-representative data.

Key dependent variables

Our main dependent variables are (1) a binary indicator of whether a respondent aged 15 years or older reports involvement in at least one surveilling institution and, among those who report any such involvement, (2) an estimate of how much time (in minutes) Latino respondents aged 15 years or older spend involved in surveilling institutions.Footnote ⁹ The first reflects the incidence and the second reflects the conditional degree of reported institutional involvement.Footnote ¹⁰

We construct our outcomes for different racial/ethnic groups, as guided by our theoretical framework. First, we measure the incidence and conditional degree of institutional involvement for all Latino ATUS respondents, regardless of nativity and citizenship status. We do so because, in a racialized system of immigration policing, U.S.- and foreign-born Latinos alike report fearing deportation (Abrego Reference Abrego2019; Asad Reference Asad2020; Flores and Schachter Reference Flores and Schachter2018; Menjívar Reference Menjívar2021). Second, we construct our outcome measures for Latinos by nativity (i.e., U.S. or foreign born) and citizenship status (i.e., naturalized citizens or noncitizens). Latinos across nativity and citizenship statuses report fearing deportation, but these reports are highest among noncitizens – all of whom are vulnerable to deportation (Asad Reference Asad2020; Johnson et al. Reference Johnson, Levesque, Lewis and Asad2024; Lai et al. Reference Lai, Hoffmann and Waldinger2023; Waldinger et al. Reference Waldinger, Hoffmann and Lai2023). The “noncitizen” category encompasses permanent residents, visa holders, and undocumented immigrants, the last of which is disproportionately vulnerable to deportation. Although the ATUS data do not report noncitizens’ specific legal statuses, any identified relationship between immigration policing and institutional involvement across state-years should be conservative since undocumented immigrants are likely more responsive to the threat of immigration policing.Footnote ¹¹ Finally, as a falsification test, we measure the incidence and conditional degree of U.S.-born White citizens’ reported involvement in surveilling institutions, a group for whom we have less expectation of a relationship between the spatial and temporal context of immigration policing and system involvement.

The time-diary data allow us to measure institutional involvement among ATUS respondents. For our binary indicator of the incidence of involvement, we code as “1” respondents who reported spending at least one minute engaged with a medical, financial, educational, and/or labor market institution in the 24 hours preceding the survey. This operationalization follows prior research (Brayne Reference Brayne2014; Desai et al. Reference Desai, Jessica Houston and Adelman2020; Haskins and Jacobsen Reference Haskins and Jacobsen2017; Lanuza and Turney Reference Lanuza and Turney2020).Footnote ¹² For our estimate of the conditional degree of involvement, we sum the number of minutes each respondent reported involved in surveilling institutions among those who reported any such involvement in the 24 hours preceding the survey. In the ATUS data, involvement in medical institutions refers to the use or purchase of medical and care services for oneself or any children in one’s care; these can encompass having or paying for a health-care appointment, as well as talking with a health-care professional, but does not include “self-care” activities (e.g., haircuts, facials, pedicures). Involvement in financial institutions includes the use or purchase of financial services and banking for oneself; examples range from applying for a credit card or loan to doing banking (e.g., interacting with an online bank account, going to a bank in person, meeting with a bank teller). Involvement in educational institutions includes participation in educational activities for oneself or any children in one’s care; activities must be non-work related, such as taking classes, conducting research, administrative tasks, and extracurricular activities (except sports). Finally, involvement in labor market institutions includes participation in formal and informal work and work-related activities for oneself; activities include being at work (and any associated tasks that result in payment) and preparing goods for sale (e.g., food, drink, pottery, etc.). A review of research using these indicators notes their construct validity and external reliability (Hamermesh et al. Reference Hamermesh, Frazis and Stewart2005). Though our main interest is in the incidence and conditional degree of involvement in surveilling institutions overall, we also examine results by institutional type.

Our analytical samples include respondents aged 15 or older who identify as Latino and who completed a time diary. After dropping respondents with missing data, our final samples are the following: all Latinos (18,033 observations), U.S.-born Latinos (8,783 observations), foreign-born Latinos (9,250 observations), Latino naturalized citizens (3,433 observations), and Latino noncitizens (5,817 observations). For our falsification test, we rely on a sample of U.S.-born White respondents aged 15 or older who completed a time diary (82,417 observations).

Key independent variables

A state-year measure of the logged ICE detainer rate is our primary independent variable. We construct the rate by taking the number of detainers per 1,000 noncitizens in a state-year. We scaled the number of detainers to the noncitizen population to account for the size of the population at risk of immigration policing. The rate’s distribution is right-skewed, which we account for with the log transformation. We focus on the state-year because, as we show in Appendix Table A2, spatial and temporal dynamics explain more variation in the logged ICE detainer rate (R ² = 0.811) than state (R ² = 0.592) or year fixed effects (R ² = 0.218) separately.

Although detainers capture one aspect of state and local police collaboration with federal immigration authorities, what detainers mean to populations worried about deportation likely depends on other structural aspects of immigration policing that can also vary across space and time. We focus specifically on the presence or absence of a sanctuary policy that limits at least some police cooperation with ICE in a state-year. We create a binary indicator that is coded as “1” when a state refuses to honor at least some ICE detainer requests. We first relied on Hausman’s (Reference Hausman2020) publicly available database on sanctuary policies between 2009 and 2018. We consulted Hausman’s source material to validate these data, which he assembled for the largest U.S. counties by Latino population according to the 2010 U.S. Census. We updated the database to include more years and geographies by relying on the source material cited in Lasch et al. (Reference Lasch, Chan, Eagly, Francesca Haynes, Lai, McCormick and Stumpf2018). Within our analytical sample of “All Latinos,” about 18% lived in a state-year with a sanctuary policy between 2007 and 2018.

Finally, beyond the threat of immigration policing, individuals worried about deportation may face differential access to specific surveilling institutions depending on other laws, policies, or regulations they must manage across state-years. Given the results of our main analysis, we consider whether two of these moderate our observed associations between immigration policing and the incidence and conditional degree of involvement in work or educational institutions. For involvement in work institutions, we consider whether a state-year participates in E-Verify, a federal program that requires public and/or private employers to verify the work authorization of their employees. We code any such participation as “E-Verify” and no such participation as “No E-Verify.” Finally, for involvement in educational institutions, we consider whether a state-year offers in-state tuition benefits for undocumented immigrants. “Comprehensive” refers to in-state tuition benefits available at all universities and colleges in a state-year; “Partial” refers to these benefits at some universities and colleges in a state-year; and “None” refers to these benefits being unavailable at all universities and colleges in a state-year.

All preceding variables are observed 1 year prior to the ATUS survey year to minimize the possibility that reverse causality accounts for our results.

Control variables

We adjust our models for state-level characteristics that may be correlated with institutional involvement based on the literature reviewed. As with the immigration policing measures, most time-varying control variables are observed 1 year prior to the year of the ATUS survey (exceptions indicated below). Appendix Table A1 details these indicators’ construction. Table 1 summarizes their mean, standard error, minima, and maxima.

Table 1.

Summary Statistics for Outcome, Explanatory, and Control Variables for Latinos in 2008–2019 American Time Use Survey

Note: N = 18,033 Latinos. The conditional average time Latinos spend involved in institutions includes travel to and from those institutions. The explanatory and control variables are the observed values in the year prior to the corresponding American Time Use Survey (ATUS) year. The mean of each variable is adjusted for survey year except for the predictability measure. Per capita police and fire protection expenses were adjusted to 2018 dollars and include both state and local-level funding. The following states were excluded from the analyses: Alaska, Hawaii, Montana, New Hampshire, South Dakota, Vermont, West Virginia, and Wyoming.

We control for individual- and state-level demographic and sociolegal contexts. At the individual level, we control for respondents’ gender, ethnic origins, age, years of residence in the United States, educational attainment, marital status, household income, the number of household adults, whether they are parents, employment status, school enrollment status, and whether they completed the ATUS interview in English.

At the state level, we measure demographic and sociolegal contexts. Our state-year demographic measures include the total population (in thousands) and the percentage of the population that is Latino. Our state-year sociolegal measures include the percentage of the state that voted Republican in the most recent presidential election year, the per capita police and fire protection expenses in a state in 2018 dollars, and a state’s total crime rate per capita. We also include state and year fixed effects in all analyses.Footnote ¹³

Analytic strategy

Our goal is to examine whether immigration policing across state-years is associated with Latinos’ reported involvement in surveilling institutions. Given our hypotheses that immigration policing matters for the incidence but not conditional degree of Latinos’ reported involvement in surveilling institutions, we use a double-hurdle model.Footnote ¹⁴ A double-hurdle model assumes that participation in an activity results from two processes: whether to participate at all and, among participants, how much to participate (see, e.g., Hamermesh and Trejo Reference Hamermesh and Trejo2013). Although there are different statistical models that can account for data with many zeros (here, signifying noninvolvement), the double-hurdle model is the most general (Cragg Reference Cragg1971).Footnote ¹⁵

In Hurdle 1, we estimate a complementary log-log model to predict whether a Latino respondent reports any involvement in surveilling institutions (a binary outcome of incidence). In Hurdle 2, we estimate a zero-truncated Poisson model to examine the number of minutes Latinos reported spending involved in surveilling institutions, conditional on any such involvement (a continuous outcome). Formally, the Hurdle 1 and Hurdle 2 models are as follows:

(1)\begin{equation}\begin{gathered} y_{i1}^* = {x_i}\alpha + \,{u_i} \hfill \\ {y_{i1}} = \;\left\{ {\begin{array}{*{20}{c}} {1\,\,\,\,\,\,\,if\,y_{i1}^* \gt 0} \\ {0\,\,\,\,\,\,\,if\,y_{i1}^* \leqslant 0} \end{array}} \right\}\;\;\;\;Incidence{\text{ }}of{\text{ }}Institutional{\text{ }}Involvement \hfill \\ \end{gathered} \end{equation}

(2)\begin{equation} \begin{array}{ll} y_{i2}^* = {x_i}\beta + \,{v_{\,i}} & \\[4pt] {y_{i2}} = \left\{ {\begin{array}{cl} {x_i}\beta + {v_i} & if\,y_{i2}^* \gt 0\ and\ y_{i1}^* \gt 0 \\ 0 & otherwise \end{array}} \right\} & \text{Conditional Degree of} \\[-10pt] & \text{Institutional Involvement} \end{array} \end{equation}

where $y_{i1}^*$ is a latent endogenous dichotomous variable indicating whether a Latino respondent reports any involvement in surveilling institutions, $y_{i2}^*$ is a latent endogenous continuous variable representing the degree of this involvement, ${y_i}$ is the observed outcome variable, ${x_i}$ is a vector of our explanatory and control variables, and ${u_i}$ and ${v_i}$ are error terms. We assume that the two equations are independent and, therefore, that the error terms are randomly and independently distributed. We also assume that the error terms are heteroscedastic across state-years. All models include year and state fixed effects to correct for heteroscedasticity.

Given our expectation that immigration policing has differential implications for the incidence and conditional degree of reported involvement in surveilling institutions across state-years, we present Hurdle 1 and Hurdle 2 results separately. For each hurdle, we examine the relationship between immigration policing and reported involvement in surveilling institutions for Latinos overall, by nativity and citizenship status, and by institutional type. We then present results from our falsification test with U.S.-born White ATUS respondents. Next, we analyze whether the relationship between immigration policing and the incidence and conditional degree of reported involvement in surveilling institutions across state-years varies by the presence of a sanctuary policy. Finally, and given the results of the main analysis, we examine whether laws that enable or constrain noncitizens’ access to labor market (i.e., E-Verify) or educational (i.e., in-state tuition benefits) institutions moderate our observed relationships for Latinos across state-years.

Coefficients resulting from double-hurdle models have no simple interpretation (Zaiceva and Zimmermann Reference Zaiceva and Zimmermann2014: p. 92). We, therefore, calculate, present, and interpret differences in average marginal effects (see Long and Mustillo Reference Long and Mustillo2021; Mood Reference Mood2010). In our main analysis, this approach allows us to assess if the average marginal effect of immigration policing is associated with the incidence and conditional degree of Latinos’ reported involvement in surveilling institutions across state-years. In our analysis of sanctuary policies and policies governing noncitizens’ access to surveilling institutions, this approach assesses if the average marginal effect of immigration policing on the incidence and conditional degree of reported involvement across state-years varies by our operationalizations of the different indicators discussed above.