Deportation threat predicts Latino US citizens and noncitizens’ psychological distress, 2011 to 2018

Significance

Between 2011 and 2018, psychological distress increased among Latinos overall. Latino US-born citizens and noncitizens exhibit greater, and naturalized citizens lesser, psychological distress over time. These divergent patterns only partly reflect the consequences of discrete dramatic societal events that independently signal a change to the country’s approach to deportation and/or that render deportation temporarily more salient to the public. They more consistently reflect gradual changes to the everyday institutional (i.e., the federal government’s quotidian efforts to detain and deport noncitizens) and social (i.e., deportation’s ongoing salience to a concerned public) environments of deportation threat since 2011, a dynamic that affects both Latino US citizens and noncitizens as a racialized ethnic group.

Abstract

The national context of deportation threat, defined as the federal government’s approach to deportation and/or deportation’s salience to the US public, fluctuated between 2011 and 2018. US Latinos across citizenship statuses may have experienced growing psychological distress associated with these changes, given their disproportionate personal or proximal vulnerabilities to deportation. Drawing on 8 y of public- and restricted-access data from the National Health Interview Survey (2011 to 2018), this article examines trends in psychological distress among Latinos who are US-born citizens, naturalized citizens, and noncitizens. It then seeks to explain these trends by considering two theoretical pathways through which the national context of deportation threat could distress Latinos: 1) through discrete dramatic societal events that independently signal a change to the country’s approach to deportation and/or that render deportation temporarily more salient to the public or 2) through more gradual changes to the country’s everyday institutional (i.e., quotidian efforts to detain and deport noncitizens) and social (i.e., deportation’s ongoing salience to a concerned public) environment of deportation threat. We find that, though both pathways matter to some degree, there is more consistent evidence that the gradual changes are associated with Latino US citizens and noncitizens’ overall experiences of psychological distress. The article highlights how, even absent observable spillover effects of dramatic societal events bearing on deportation threat, the institutional and social environment in which they occur implicates Latinos’ well-being.

The national context of deportation threat is defined by the federal government’s approach to deportation and/or deportation’s salience to the US public. Although experiences of this national context can vary sub-nationally (1–4), the federal government delineates deportation’s scope at both the country’s borders and throughout its interior (5–7). It determines who may enter the country, under what conditions they may remain, and the conditions under which they must exit. This latter authority has been used with increasing frequency in recent decades: of the 6.8 million deportations effected since 1986, about half occurred in the decade since 2011 (8). Whether undocumented immigrants, visa holders, or permanent residents, all noncitizens are vulnerable to deportation—with those from Mexico and Central America disproportionately impacted (9–13). Most Latino noncitizens will never experience deportation, but their shared vulnerability translates into similar deportation fears, regardless of their individual legal status (14–16). Meanwhile, Latino US citizens also fear deportation despite their ostensible immunity to it (14, 17), whether because they live with noncitizens (SI Appendix, Table S1) or because they worry that immigration officers will racialize them as noncitizens (10). The material, social, and psychological consequences of the deportation dragnet, therefore, reach Latinos who are noncitizens and US citizens (18–22).

In this article, we evaluate two theoretical pathways through which the national context of deportation threat may psychologically distress Latinos across citizenship statuses, especially those who may not be its direct targets. One pathway is through “dramatic societal events” (23), defined as discrete incidents that independently signal a change to the country’s approach to deportation and/or render deportation temporarily more salient to the public. Between 2011 and 2018, several such events occurred, reflecting a federal tug-of-war over deportation’s scope. For example, in June 2012, the Obama administration announced Deferred Action for Childhood Arrivals (DACA) to grant some noncitizens temporary reprieve from deportation. The Obama administration expanded DACA with Deferred Action for Parents of Americans and Lawful Permanent Residents (DAPA) in November 2014; DAPA was never implemented because dozens of states were granted injunctive relief in federal court almost immediately thereafter. DAPA remained blocked until after the 2016 presidential election, with the Trump administration rescinding it in June 2017. The Trump administration then sought to rescind DACA (resulting in a yearslong tussle in federal court that ultimately preserved DACA on an administrative technicality). Existing research links some dramatic societal events, most notably the 2016 presidential election, to mental health consequences for Latino noncitizens and US citizens (14, 17, 24, 25). However, we know far less about whether other dramatic societal events, such as DACA’s introduction or attempted rescission, have mental health consequences for Latinos beyond those segments eligible for these initiatives (26–30).

Another pathway is through more gradual changes to the country’s everyday institutional and social environment of deportation threat. Whereas the institutional environment refers to the federal government’s quotidian efforts to detain and deport noncitizens, the social environment refers to the salience of these efforts to a concerned public. Rarely do dramatic societal events occur in a vacuum; they instead reflect the institutional and social environments in which they occur. For example, the executive branch began prioritizing the deportation of noncitizens with “serious” criminal convictions in 2011, 1 y before announcing DACA (itself an example of deprioritizing the deportation of some noncitizens) (31, 32). As the Obama administration implemented these changes, the social environment of deportation threat may have gradually attenuated along with the institutional environment, as qualitative research suggests (1–4). These dynamics may have shifted again during the Trump administration, which both did away with deportation priorities and represented an altogether distinct social environment of deportation threat (14, 24, 25, 33). Even absent observable evidence of a dramatic societal event’s impact on mental health, then, Latinos across citizenship statuses may perceive the country’s everyday institutional and social environments of deportation threat as more or less distressing.

This article examines trends in psychological distress between 2011 and 2018 among Latinos who are US-born citizens, naturalized citizens, and noncitizens. It then considers whether two aspects of the national context of deportation threat—specifically, dramatic societal events or more gradual changes to everyday institutional and social environments—explain these trends.

Research Strategy

This article relies on public- and restricted-access data from the National Health Interview Survey (NHIS; 2011 to 2018) (34, 35). Although the survey is cross-sectional, the NHIS is nationally representative and regularly polls adults aged 18 and older to measure their healthcare access; mental and physical health; and other health behaviors. We focus on the Sample Adult portion of the NHIS, in which one adult per sampled household is selected at random by the computerized survey instrument to answer additional health-related questions, including those that allow us to measure psychological distress, our outcome of interest. The resulting sample size is over 770,000 non-institutionalized adults; our regression discontinuity design (RDD) nonetheless requires that we rely on various subsets of this sample, specified below, when analyzing the restricted-access data. As a falsification test that we summarize below and explain more fully in Materials and Methods, we also include comparisons to non-Hispanic Black and White US-born citizens.

Our dependent variable is non-specific psychological distress, which measures well-being and serves as a barometer of population mental and physical health (36–39). We observe psychological distress by relying on the Kessler-6 (K6) index, which is a well-validated instrument capturing the overall frequency of anxious and depressive symptoms over the past 30 d (36, 37). The K6 index consists of six questions focused on the emotions of sadness, nervousness, restlessness, worthlessness, and hopelessness, as well as the feeling that everything is an effort. Each question is measured on a frequency scale from 0 to 4, in which 0 indicates the respondent experiences the emotion “none of the time” and 4 indicates “all of the time.”

We measure psychological distress in four ways. First, we take the sum of the K6 items, which ranges from 0 to 24; higher scores indicate greater overall psychological distress. Second, following prior research (30, 38), we create a binary indicator of moderate or severe psychological distress for respondents whose K6 value is five or higher (1 if K6 ≥ 5; 0 otherwise). While our first and second measures examine the K6 index as a unidimensional instrument of overall psychological distress, our third and fourth measures follow a growing literature discussed in Materials and Methods and examine the index’s constituent dimensions of anxiety and depression (39–44). Our anxiety measure averages the frequency of the two anxious symptoms (i.e., nervousness and restlessness) and provides a scale indicating their overall frequency (0 = none of the time and 4 = all of the time). Our depression measure averages responses to the four depressive symptoms (i.e., sadness, worthlessness, hopelessness, and the feeling that everything is an effort) and is measured like the anxiety subscale. In our main analysis, we provide model results for the K6 score and the binary measure of moderate or severe psychological distress. In cases where we find that an event significantly affects psychological distress, we explore in explanatory analyses the underlying emotional response.

Our primary independent variable is a combined measure of race/ethnicity and citizenship. Given their disproportionate vulnerability to and targeting for deportation, we focus in our main analysis on respondents who self-identify as Latino or Hispanic of any racial identification and separate them into US-born citizens, naturalized citizens, and noncitizens. In explanatory analyses, we further divide these groupings into two broader ethnic categories: Mexican or Central American (arguably the segment of the Latino population most affected by each of the dramatic societal events considered in this article) and Other (arguably the segment least affected). The latter category includes non-Mexican or -Central American ethnicities, such as respondents who report Puerto Rican, Dominican, or Cuban ancestry. Finally, in a falsification test that we present throughout the main results, we identify non-Hispanic Black and White US-born citizens as a comparison group. As a population neither vulnerable to nor targeted for deportation and whose broader racial groups are not typically racialized as “foreign” in daily life (10, 45), we have less theoretical expectation of a relationship between the national context of deportation threat and psychological distress.^* Additional theoretical justification for this falsification test is provided in Materials and Methods.

Data analysis proceeds in two interconnected parts. First, using the publicly available data, we plot age-adjusted descriptive linear trends in psychological distress between 2011 and 2018 for the racial/ethnic and citizenship groupings specified above. We characterize these different groups’ age-adjusted trends in psychological distress as rising, stable, or decreasing over time, based on the slope of the linear trend lines (indicated by β below). We also consider statistically significant differences based on 95 percent CI in the mean value of each measure of psychological distress in 2011, 2018, and across the full study period.

Second, and based on the theoretical and empirical literature reviewed in the Introduction, we consider two possible explanations for our observed trends. One is dramatic societal events that independently signal a change to the country’s approach to deportation and/or render deportation temporarily more salient to the public. To examine this possibility, we rely on information regarding the exact date a respondent completed the NHIS. We obtained these dates via restricted-access data from the National Center for Health Statistics (NCHS) that we analyzed at the Federal Statistical Research Data Center (RDC) at Stanford University (35). The publicly -available NHIS data include only the month and year of interview completion, which do not provide sufficient specificity to capture the effects of any single event considered in this article. We use the date of survey completion to calculate event “cutoffs” indicating whether the respondent completed the Sample Adult section of the NHIS questionnaire on or after the event date. We also construct event “bandwidths” based on the date of survey completion that capture the subset of respondents who completed their survey within 30 d on either side of the event. These subsets serve as analytical samples that meet the assumptions of the RDD described below. The six events analyzed (and their date) include:

• 1.DACA’s Announcement (15 June 2012)
• 2.DAPA’s Announcement (20 November 2014)
• 3.DAPA’s Injunction (16 February 2015)
• 4.Donald Trump’s Election to the US Presidency (8 November 2016)
• 5.DAPA’s Rescission (15 June 2017)
• 6.DACA’s Rescission (5 September 2017)

We selected these events for several reasons. First, though DACA and DAPA concern a fraction of the over 10 million undocumented noncitizens in the United States, at the time of their announcement, they reflected a broader shift in the federal government’s approach to immigration enforcement toward prioritizing the deportation of noncitizens with criminal convictions (31, 32). Second, given their disproportionate vulnerability to and targeting for deportation by immigration authorities, Latino noncitizens constitute the majority of DACA and would-be DAPA recipients (46–49). Third, and coupled with the prior points, 1.7 million Latino US citizens share a household with at least one undocumented noncitizen, regardless of their eligibility for DACA and DAPA (SI Appendix, Table S1). Fourth, Donald Trump launched his campaign for the US presidency in part to dismantle enforcement priorities like DACA and DAPA (50). The theoretical relationship between each event—as independent yet interconnected moments tempering or magnifying the national context of deportation threat for Latino noncitizens and US citizens—is well suited for our analysis. We expect that, whereas DACA and DAPA’s announcements are associated with reductions in psychological distress for Latinos across the different citizenship subgroups, Donald Trump’s election and his administration’s (attempted) rescissions of the enforcement priorities are associated with increases. We further expect that any changes in psychological distress reflect changes in reported anxiety rather than depression symptoms, given anxiety’s noted responsiveness to contextual changes (39).

To identify the plausibly causal effect of the six selected dramatic societal events, we use a RDD with local linear regression.^† Prior studies suggest that RDDs approach randomized experiments in their efficacy when they meet the “continuity assumption”: assignment to the treatment is ignorable within a narrow interval around the cutoff (51, 52). We show in SI Appendix, Table S2 that NHIS respondents who completed their survey within 30 d before or after each event are comparable within each 60-d period. Still, motivated by prior research suggesting that years of US residence and English proficiency can pattern perceptions of deportation threat among Latino naturalized citizens and noncitizens (15, 16), we also present results from analyses that include these factors as control variables in SI Appendix, Table S13.

A second potential explanation is that more gradual changes to the everyday institutional and social environments of deportation threat may account for our observed trends. Although well suited for examining the effects of discrete dramatic societal events within a narrow time window, the RDD’s continuity assumption means that it cannot consider the possible mediating role of the broader institutional or social environment on psychological distress. To better situate our observed trends within these environments, we rely on two additional datasets that we merge with the publicly available NHIS data. These datasets offer us two additional independent variables. The first measures the institutional environment, which we define as the federal government’s quotidian efforts to detain and potentially deport noncitizens nationwide. From Syracuse University’s Transactional Records Access Clearinghouse (TRAC) data, we observe the number of detainer requests that the Department of Homeland Security (DHS) issued in each month-year between 2011 and 2018 (53). Detainers are requests that the DHS, within which Immigration and Customs Enforcement (ICE) is housed, makes of local police when those officers arrest and fingerprint someone whom DHS databases then suggest is deportable. Local police need not cooperate, and most detainers do not result in deportation. They are nonetheless an indicator that ICE, a federal agency, has an interest in deporting someone whom local police have arrested. To be sure, more than half of all recent interior (rather than border) deportations stemmed from a detainer request (31, 32). Our standardized month-year count of detainers, therefore, reflects the institutional environment of deportation threat inside the country between 2011 and 2018.

The second dataset allows for a national measure of the social environment of deportation threat. From the publicly available Google Trends data, we generate a measure of the relative prevalence of Google searches nationwide for the “Immigration” topic in each month-year between 2011 and 2018 (54). While our measure of detainers reflects observable changes in the federal government’s efforts at interior enforcement, as explained more fully in Materials and Methods, our measure of Google Trends reflects observable changes in the US public’s awareness of or interest in immigration-related issues (55). Importantly, Google Trends is not a measure of attitudes. However, given that Google Trends scores are calculated based on millions of users’ search behavior, it is a reliable measure for revealing meaningful social patterns that “correlate strongly with demographics of those one might most expect to perform the searches” (56). As with the detainers measure, we standardize Google Trends scores in all analyses.

To examine the relative associations of the country’s everyday institutional and social environments of deportation threat with psychological distress, we use the publicly available NHIS to fit ordinary least squares (OLS) regressions for the three continuous measures of psychological distress (i.e., K6, anxiety, and depression) and a linear probability model (LPM) for the binary measure (i.e., moderate or severe psychological distress).

We more fully describe the data and methods, including their assumptions, limitations, and additional analyses, in Materials and Methods.

Results

Results show that 1) psychological distress among Latinos overall and non-Hispanic Black and White US-born citizens is similar and increasing across the study period; 2) citizenship status is an important dimension patterning psychological distress within the Latino population; 3) a few of the dramatic societal events considered are associated with Latino noncitizens and some US citizens’ psychological distress, and we do not observe equivalent discontinuities for non-Hispanic Black and White US-born citizens; and 4) trends in Latino US citizens and noncitizens’ psychological distress consistently track gradual changes to the country’s everyday institutional and social environments of deportation threat.

Trends in Psychological Distress, 2011 to 2018.

Fig. 1 displays US Latinos and non-Hispanic Black and White Americans’ age-adjusted monthly average rates of psychological distress (Fig. 1A), moderate or severe psychological distress (Fig. 1B), anxiety (Fig. 1C), and depression (Fig. 1D) between 2011 and 2018 using the publicly -available NHIS. We present linear trends in these figures but plot spline-based trends in SI Appendix, Fig. S1 that further substantiate our interpretations.

Fig. 1.

Age-adjusted monthly trends in (A) K6 scores, (B) moderate or severe psychological distress, (C) depressive symptoms, and (D) anxious symptoms for US Latinos by citizenship status and for non-Hispanic Black and White US-born citizens, 2011 to 2018. Source: NHIS 2011 to 2018. Note: *P < 0.05 **P < 0.01 ***P < 0.001. P-values indicate a significant change in outcome at the time of event. Models compare linear trends 30 d pre- and post-event. Psychological distress is measured using the K6 index, which varies between 0 and 24 with higher scores indicating greater distress.

Fig. 1A shows that, overall, Latinos’ age-adjusted average K6 score is 2.632 [2.567, 2.697] (95% CI; exceptions indicated) within the study period. Latinos’ average psychological distress score trends toward an increase across time (β = 0.002 [0.002, 0.002]), from 2.574 [2.423, 2.725] in 2011 to a high of 2.835 [2.633, 3.037] in 2018. These trends are similar among non-Hispanic Black and White US-born citizens (β = 0.006 [0.006, 0.006]), whose average psychological distress ranges from 2.580 [2.444, 2.572] in 2011 to 2.916 [2.830, 3.003] in 2018. Such similarities imply that citizenship status is an important axis of variation patterning psychological distress within the Latino population. Among Latinos, the US born have the highest average K6 score across the study period (2.793 [2.687, 2.900]), followed by noncitizens (2.536 [2.417, 2.655]) and naturalized citizens (2.518 [2.403, 2.632]). Latinos who are US born (β = 0.002 [0.002, 0.002]) and noncitizens (β = 0.001 [0.001, 0.001]) report increasing, and those who are naturalized citizens report decreasing (β = −0.001 [−0.002, −0.001]), K6 scores across the study period.

Fig. 1B characterizes trends in moderate or severe psychological distress. Although average rates of moderate or severe psychological distress for Latinos overall (0.211 [0.205, 0.217]) are higher than those of non-Hispanic Black and White US-born citizens (0.203 [0.200, 0.206]) across the study period, these differences are substantively small because trends in moderate or severe psychological distress for both groups converge over time (due to all groups’ increased reports of anxiety symptoms, as shown below). Within the Latino population, rates of moderate or severe psychological distress are more or less stable between 2011 and 2018 for US-born Latinos (β = 0.000 [0.000, 0.000]), naturalized citizens (β = −0.000 [−0.000, −0.000]), and noncitizens (β = 0.000 [0.000, 0.000]).

Fig. 1C demonstrates the average prevalence of reported anxiety symptoms. Non-Hispanic Black and White US-born citizens (0.652 [0.644, 0.660]) report a higher average prevalence of anxiety symptoms than Latinos overall (0.562 [0.549, 0.575]) across the study period. Both non-Hispanic Black and White US-born citizens (β = 0.002 [0.002, 0.002]) and Latinos overall (β = 0.001 [0.001, 0.001]) report increasing anxiety symptoms. Within the Latino population, the US born have the highest (0.633 [0.612, 0.655]) average prevalence of reported anxiety symptoms across the study period, followed by naturalized citizens (0.530 [0.506, 0.553]) and noncitizens (0.499 [0.476, 0.522]). US-born Latinos (β = 0.001 [0.001, 0.001]) exhibit similar trends in reported anxiety symptoms as non-Hispanic Black and White US-born citizens. Latino noncitizens’ reported anxiety symptoms are also ticking up across time (β = 0.001 [0.001, 0.001]), but those of Latino naturalized citizens are stable (β = 0.000 [0.000, 0.000]).

In Fig. 1D, we observe the average prevalence of reported depression symptoms. Latinos overall (0.377 [0.366, 0.388]) report a higher average prevalence of depressive symptoms than non-Hispanic Black and White US-born citizens (0.329 [0.323, 0.334]), though prevalence increases across the study period for the latter group (β = 0.001 [0.001, 0.001]). Within the Latino population, average reports of depressive symptoms do not vary among US-born citizens (0.382 [0.364, 0.399]), naturalized citizens (0.365 [0.345, 0.383]), and noncitizens (0.384 [0.364, 0.404]). These rates are largely stable between 2011 and 2018 for the US born (β = 0.000 [0.000, 0.000]), naturalized citizens (β = −0.000 [−0.000, −0.000]), and noncitizens (β = −0.000 [−0.000, −0.000]).

These trends suggest that, while psychological distress is increasing among Latinos overall and non-Hispanic Black and White US-born citizens across the study period, Latino US-born citizens, naturalized citizens, and noncitizens report distinct psychological distress experiences.

Dramatic Societal Events.

One explanation for these trends is that dramatic societal events affect psychological distress among Latino US-born citizens, naturalized citizens, and noncitizens by independently signaling a change to the country’s approach to deportation regarding some noncitizens and/or rendering deportation temporarily more salient to the public, even for those not directly targeted.

Fig. 2 presents the RDD results analyzing the restricted-access NHIS data and examines the effect of different dramatic societal events on psychological distress for the distinct citizenship groupings of Latinos, as well as non-Hispanic Black and White US-born citizens. Results from the RDD match the descriptive statistics presented in SI Appendix, Table S2. We generally observe no spillover effect of any event considered on the predicted average K6 score for any Latino subgroup, with two important exceptions. First, among Latino naturalized citizens, we observe decreases in average K6 scores following DACA (−2.809 [−5.247, −0.371] and DAPA’s (−2.431 [−4.469, −0.393]; 90%CI) announcements—but only after controlling for years of US residence and interview language (SI Appendix, Table S13).^‡ Second, among Latino noncitizens, we find a significant increase in the predicted average K6 score (2.003 [0.343, 3.664]) immediately following the 2016 presidential election. This reflects an increase in both anxious (0.381 [0.076, 0.686]) and depressive (0.310 [0.012, 0.609]) symptoms, as shown in the SI Appendix, Fig. S2. We show in SI Appendix, Fig. S3 that the psychological effect of the 2016 presidential election is similar among Latino noncitizens from Mexico and Central America and from elsewhere in Latin America. We observe no equivalent discontinuities for non-Hispanic Black and White US-born citizens’ K6 scores, a result that is substantively similar when examining each racial group separately (SI Appendix, Fig. S4) and when considering reported anxious and depressive symptoms (SI Appendix, Fig. S2).^§

Fig. 2.

Regression discontinuity of psychological distress for US Latinos by citizenship status and non-Hispanic Black and White US-born citizens following six dramatic societal events. Source: NHIS 2011 to 2018. Note: ⁺P < 0.1 *P < 0.05 **P < 0.01 ***P < 0.001. P-values indicate a significant change in outcome at the time of event. Models compare linear trends 30 d pre- and post-event. Severe psychological distress indicates a K6 score ≥ 5.

Fig. 3 is like Fig. 2 but focuses on reports of moderate or severe psychological distress among the different Latino subgroups and non-Hispanic Black and White US-born citizens in the restricted-access NHIS data (SI Appendix, Table S3). First, we do not observe that the increases in Latino noncitizens’ predicted K6 scores following the 2016 presidential election from Fig. 2 translate into elevated rates of moderate or severe psychological distress. Second, we find evidence that Latino naturalized citizens also trend toward discontinuities in psychological distress following two dramatic societal events. After the announcement of DAPA on 20 November 2014, we observe that the predicted proportion of Latino naturalized citizens reporting moderate or severe psychological distress decreases by 0.237 (−0.446, −0.028; 90% CI). As we show in SI Appendix, Fig. S2, this result reflects a marginal decrease in depressive symptoms (−0.350 [−0.697, −0.003]; 90% CI). Following the 2016 presidential election, we find that the predicted proportion of Latino naturalized citizens reporting moderate or severe psychological distress trended toward an increase (0.256 [0.021, 0.491]; 90% CI). As we show in SI Appendix, Fig. S2, this worsening of distress is driven by an increase in this group’s reported anxious symptoms (0.609 [0.080, 1.137]). Although the changes in naturalized citizens’ K6 scores following DAPA’s announcement and the 2016 presidential election are not large enough to be statistically meaningful (Fig. 2), attention to the binary indicator shows how even small changes in the K6 score may implicate experiences of moderate or severe psychological distress. We do not observe any equivalent discontinuities in moderate or severe psychological distress among non-Hispanic Black and White US-born citizens for any of the dramatic societal events considered, a result that is substantively similar when examining each racial group separately (SI Appendix, Fig. S5).

Fig. 3.

Regression discontinuity of moderate or severe psychological distress for US Latinos by citizenship status and non-Hispanic Black and White US-born citizens following six dramatic societal events. Source: NHIS 2011 to 2018. Note: 95% CI produced with robust SE clustered at the month-year level. Models include controls for gender, age, employment status, household income, parenthood, marital status, education, census region of residence, and interview language, year fixed effects, and month fixed effects.

Everyday Institutional and Social Environments.

While the dramatic societal events considered are important, their effects on psychological distress may emerge from the country’s everyday institutional and social environments of deportation threat. A second explanation for our observed trends in psychological distress, then, is that they reflect more gradual changes to the federal government’s quotidian interior enforcement actions (SI Appendix, Fig. S6), these actions’ ongoing salience to a concerned US public (SI Appendix, Fig. S7), or both.

Fig. 4 tests the association of the institutional and social environments of deportation threat with psychological distress using the publicly available NHIS data. We present regression tables in SI Appendix, Tables S4 and S5. For US-born Latinos, increases in the Google Trends, but not the detainers, measure is associated with increases in all four measures of psychological distress, all else equal. A one SD increase in the Google Trends measure is associated with a 0.089 (0.018, 0.159) unit increase in K6 score, a 0.008 (0.001, 0.015; 90% CI) increase in the predicted proportion reporting moderate or severe psychological distress, a 0.022 (0.006, 0.039) unit increase in anxious symptoms, and a 0.011 (0.000, 0.022) unit increase in depressive symptoms. The reverse is true for Latino noncitizens, for whom a one SD increase in detainers, but not the Google Trends, measure is associated with increases in their K6 score (0.626 [0.145, 1.106]), moderate or severe psychological distress (0.049 [0.004, 0.094]), and depressive symptoms (0.126 [0.046, 0.206]). Among Latino naturalized citizens, we observe that increases in the standardized detainer count are associated with decreases in reported depressive symptoms (-0.082 [-0.158, -0.006]; 90% CI). We show in SI Appendix, Table S6 that this pattern varies by Latino naturalized citizens’ socioeconomic status, with those below the poverty line reporting stronger relationships between detainers and K6 scores (0.304 [0.022, 0.585]), as well as depressive symptoms (0.057 [0.012, 0.102]), than those above the poverty threshold. Results for non-Hispanic Black and White US-born citizens, whether considered as a group or separately, are oppositely signed and/or not statistically significant (SI Appendix, Table S5).

Fig. 4.

Regression of psychological distress on ICE detainers and Google trends score for immigration topic for US Latinos by citizenship status and non-Hispanic Black and White US-born citizens, 2011 to 2018.

Discussion

This article uses 8 y of population-representative survey data to examine whether and how changes to the national context of deportation threat have spillover effects on US Latinos’ psychological distress. As a group, Latinos exhibit increasing average reports of psychological distress between 2011 and 2018. This pattern nonetheless varies by Latinos’ citizenship status. Latino noncitizens encompass undocumented and multiple categories of documented immigrants; they report an increasing average incidence of psychological distress across the study period. Among US citizens, Latino naturalized citizens differ from the US born in their experiences of psychological distress. While naturalized citizens’ average reports of psychological distress are falling across the study period, US-born citizens’ are increasing. That these trends differ from those of non-Hispanic Black and White US-born citizens supports the idea that citizenship status is an important axis of difference for evaluating whether and how Latinos experience psychological distress.

We considered two possible explanations for these observed trends. For each citizenship subgroup of Latinos, and with few exceptions that we explicate below, we find modest support that any one dramatic societal event examined accounts for these results. We instead find that gradual changes to the country’s everyday institutional and social environments of deportation threat, specifically to the federal government’s quotidian efforts to detain noncitizens for possible deportation and to deportation’s ongoing salience to a concerned US public, explain them. The institutional and social environments affect Latinos who are US born, naturalized citizens, and noncitizens through separate but related pathways. Whereas deportation’s greater public salience is associated with increases in psychological distress among the US born, intensified enforcement actions are associated with increases in psychological distress among Latino noncitizens overall and the most socioeconomically disadvantaged naturalized citizens. Overall, these results uncover how, even absent observable spillover effects of a dramatic societal event, the country’s everyday institutional and social environments of deportation threat have far-reaching consequences for Latinos’ well-being.

Additional work is needed to extend the analyses presented here. First, although we measured psychological distress as both a one- and two-dimensional instrument, research attuned to other outcomes may uncover more insights into how the national context of deportation threat is associated with well-being. Some examples include perceived discrimination and infant health (57, 58). Second, measures that allow for the disaggregation of noncitizens into their constituent legal statuses (i.e., lawful permanent resident or otherwise) would facilitate greater insight into whether they experience the national context of deportation threat differently. The NHIS does not collect these data, though strategies that impute legal status may prove fruitful (59). Third, although differential proximity of the Latino citizenship subgroups to undocumented immigrants is unlikely to explain our results (SI Appendix, Table S1), information on the citizenship and/or legal status of respondents’ household members could more directly test this idea. For example, permanent residents married to undocumented immigrants are less likely than permanent residents married to naturalized citizens to become citizens (60); noncitizens in the former dynamic may be more responsive to the national context of deportation threat. Fourth, although the NHIS data illuminate important cross-sectional nuances in Latinos’ psychological distress over time, longitudinal and survey experimental data are needed to illuminate the precise mechanisms linking the national context of deportation threat and psychological distress. This research also can consider whether and how subnational contexts matter. Finally, comparisons across racial/ethnic groups racialized as foreign are in order. As the fastest-growing foreign-born group in the United States, attention to psychological distress among Asians across citizenship statuses may reveal the understudied effects of deportation threat on this group (61).

Results from this study nonetheless contribute to research on the effects of a changing national context of deportation threat on the well-being of Latinos across citizenship statuses. Research suggests that Latino noncitizens are hyperaware of changes to federal immigration laws, regulations, and policies (62–64). Such hyperawareness means that they may perceive different dramatic societal events implicating deportation threat as business as usual; they may understand that different events disproportionately threaten them but see these events as constituting contemporary life in the United States. Results from this study support this characterization. Though Latino noncitizens report a lower average incidence of psychological distress relative to US-born Latinos, that incidence may be ticking up due to higher average reports of anxiety that are associated with changes to interior enforcement actions between 2011 and 2018 (65). These trends precede the 2016 presidential election; yet, as a pivotal moment marking changes to the institutional and social environment of deportation threat, the election did have an outsize impact on Latino noncitizens’ reports of psychological distress (24, 25, 66).

The divergent trends among Latino US citizens are also notable. At first blush, it may seem odd that naturalized and US-born citizens report distinct experiences of psychological distress, given that their shared citizenship ostensibly immunizes them from deportation threat (67). Yet, a growing body of research offers important context for this result. For naturalized citizens, each dramatic societal event may remind them of their own precarious position in the time prior to their naturalization (68–71). The reductions in psychological distress that Latino naturalized citizens experienced following DACA and DAPA’s announcements, especially among those who completed their surveys in Spanish, may have stemmed from the hope that some undocumented immigrants would likewise find greater stability in the country. By contrast, Donald Trump’s election to the presidency in 2016 may have reminded some Latino naturalized citizens of not just undocumented immigrants’ precarity but also their own in light of campaign speeches that framed Latino immigrants generally as detrimental to society (14, 24, 25). Such perceived precarity may also explain why we observe a relationship between detainer requests and psychological distress among the most disadvantaged Latino naturalized citizens: As the country’s institutional environment of deportation threat becomes more punitive, only the most privileged naturalized citizens may view their citizenship status as a psychological safeguard.

For US-born Latinos, we do not observe that any of the dramatic societal events considered generate discontinuities in psychological distress. Yet, changes to the social environment of deportation threat (i.e., its ongoing salience to the public) are associated with average increases in US-born Latinos’ reports of psychological distress, net of a range of controls and month and year fixed effects. In this way, though US-born Latinos may not perceive any single dramatic societal event examined as sufficiently threatening to their collective well-being, they may experience more chronic forms of anxiety and stress associated with the ebbs and flows of the national conversation surrounding immigration enforcement (72). This may occur for many reasons, such as if these individuals experience greater discrimination on the basis of their racialized legal status or experience greater stress and conflict in their household owing to immigration enforcement (10, 73–75). What citizenship means for well-being, then, may increasingly depend on Latinos’ relationship to immigration enforcement—whether as direct or indirect targets of enforcement actions or essentializing racist invectives.

The distinct impact of our measures of the everyday institutional and social environments of deportation threat on the psychological distress of US-born Latinos and Latino noncitizens also merits mention. Some of this may reflect differences by Latinos’ citizenship status in access to the internet (see SI Appendix, Table S7 for all tabulations in this paragraph). While 94 percent of Latinos overall have internet access, that rate is 97 percent among the US born and 88 percent among the foreign born; among the foreign born, access is likely highest among naturalized citizens and lowest among noncitizens (76). Although 90 percent of Latinos overall have internet access and a laptop or desktop, smartphone, or tablet, that rate is 95 percent among the US born and 87 percent among the foreign born. What’s more, 86 percent of Latinos own smartphones, a rate comparable to non-Hispanic Black and White people in the country; among Latino immigrants, this rate is 75 percent. Latinos are more likely than White or Black US-born citizens to report being dependent on their cell phones (77). This dependence is highest among Latinos who are younger, lower-income, and with a high school education or less—all characteristics correlated highly with the Latino noncitizen population (77). Cell phone dependence, in turn, translates into Latinos, especially noncitizens, being far more likely (46 percent) than Black (23 percent) or White (16 percent) Americans to report communicating regularly via WhatsApp (78).

What might these differences in technology access and use reveal about our observed results? It may be that, at least nationally, Latino US-born citizens are more likely to seek out information related to immigration enforcement via Google searches than are Latino noncitizens. This would explain why the Google Trends measure predicts US-born Latinos’ psychological distress but not Latino noncitizens’. At the same time, the association of detainers with Latino noncitizens’—but not US-born Latinos’—psychological distress may reflect one or two complementary realities: either noncitizens are more attuned to national changes in detainers issued or detainers proxy this group’s efforts to share information about interior immigration enforcement actions nationwide via social media channels like WhatsApp, as qualitative research suggests (79, 80). Taken together, we believe that the distinct impact of our multiple measures urges greater scholarly attention to both institutional and social environments of deportation threat. While the former may be measured with data on policies and practices that make deportation more or less likely, the latter may be measured with search or social media data (55, 81).

Whatever the processes underlying the results, the study points to the spillover consequences of a changing national context of deportation threat on Latino US citizens and noncitizens. Although individual dramatic societal events may themselves impact Latinos’ well-being, focusing on them in isolation risks overlooking the forest for the trees. Even when different dramatic societal events may not be experienced as independently distressing moments, they may still reflect more gradual but fundamental shifts in the country’s everyday institutional and social environment of deportation threat that have consequences for Latino US citizens and noncitizens’ well-being.

Materials and Methods

We rely on public- and restricted-access NHIS data between 2011 and 2018. The NHIS is a cross-sectional household interview survey focused on health outcomes that is representative of the US non-institutionalized population. We use the Sample Adult portion, which includes additional questions answered by a randomly selected adult aged 18 and older within each family. All analyses, including the descriptive statistics presented in the SI Appendix, are weighted using Sample Person weights adjusted for pooled survey years. The full sample size is over 770,000 respondents, though we rely on various subsets of this sample in our RDD analyses; subset sample sizes are noted in SI Appendix.

We accessed public-use NHIS data via the Integrated Public Use Microdata Series (IPUMS) at the University of Minnesota (34). The public-use data allow us to produce time trends at the month-year level (i.e., Fig. 1) and to test our claim that the country’s everyday institutional and social environments of deportation threat in part underlie these observed trends (i.e., Fig. 4). For analyses of the effects of dramatic societal events, we rely on the restricted-access NHIS (35). The third and corresponding authors submitted a research proposal describing the analyses in this article to the NCHS, which assessed the proposed analyses for possible disclosure risk. Upon approval, the corresponding author applied for and obtained Special Sworn Status to analyze these data in the RDC at Stanford University. These data contain the indicator of the exact date a respondent completed their survey. We use this date to calculate cutoffs for each societal event indicating whether the respondent completed their interview on or after the event date. As described below, and to facilitate our RDD, we also construct event bandwidths based on the date of survey completion indicating the subset of respondents who completed their survey close to the event date. Analyses were performed with Stata 15/16.

Our dependent variable is nonspecific psychological distress, which we observe using the K6 index. The K6 consists of six questions focused on the emotions of sadness, nervousness, restlessness, worthlessness, hopelessness, and the feeling that everything is an effort. Each question is measured on a frequency scale from 0 to 4, in which 0 indicates the respondent experiences the emotion none of the time and 4 indicates all of the time within the last 30 d. From these questions, we generate four measures of our outcome variable. The first takes the sum of the six questions, which ranges from 0 to 24; higher K6 scores indicate greater psychological distress. The second is a binary measure of moderate or worse psychological distress, which equals 1 when the sum of the K6 score is at least 5. Both these measures align with the K6 index’s original conceptualization as a unidimensional instrument that captures a single overall factor of non-specific psychological distress (36, 37). Much recent research nonetheless validates the K6 index as capturing two factors of psychological distress: anxiety and depression (39–44). This literature not only establishes the validity of the K6 index’s two-dimensional structure for separately measuring anxious and depressive symptoms but also demonstrates that anxiety and depression represent distinct emotional experiences—with anxiety more responsive than depression to contextual changes that may be perceived as threatening (39). To accommodate this in our analysis, we averaged the frequency of the two anxious symptoms (i.e., nervousness and restlessness) and the four depressive symptoms (i.e., sadness, worthlessness, hopelessness, and the feeling that everything is an effort) in the K6 index. The resulting scales range from 0 to 4 and indicate the overall frequency of anxious or depressive symptoms. Both measures have high construct validity, comparable to the full K6 index (α_k6 = 0.86; α_anxiety = 0.75, α_depression = 0.83). In addition, we use factor analysis to evaluate the subscales’ dimensionality. We find that, across the measures for each subscale, there is either only one valid component with an eigenvalue near (for the anxiety subscale) or above 1 (for the depression subscale) (SI Appendix, Table S8). Based on the Kaiser rule (82), this suggests that there is a single, underlying latent variable for each set of measures. We rely on the K6 index as a unidimensional instrument for our primary analyses, reserving our analyses of its two-dimensional structure to explain the underlying emotional response of any observed associations.

Our primary independent variable measures NHIS respondents’ race/ethnicity and citizenship. We focus on respondents who self-identify as Latino or Hispanic of any race and separate them into US-born citizens, naturalized citizens, and noncitizens. Although information on noncitizens’ specific legal status would uncover whether our observed trends reflect the experiences of a single category of noncitizens—such as undocumented immigrants—the NHIS does not collect these data. Still, other surveys show that noncitizens—whether permanent residents or undocumented immigrants—report comparable experiences of deportation threat despite their unequal vulnerabilities (2, 14, 83). The available data on citizenship status are, therefore, well suited for our purposes. In explanatory analyses, we divide each citizenship grouping according to respondents’ reported ancestry to better reflect their distinct vulnerabilities to and experiences with immigration enforcement (12): Mexican or Central American and Other.

We plot age-adjusted, monthly trends of our measures of psychological distress over time in Fig. 1 using the public-access NHIS data. We do so for 1) all Latino respondents, regardless of citizenship; 2) Latino respondents by their citizenship status; and 3) a single grouping of non-Hispanic White and Black US-born respondents. We assess the overall mean of each measure of psychological distress for each group across the full study period, overall means for each measure in 2011 and 2018, and the slopes of bivariate regression models predicting each measure using the month-year of the NHIS interview. We use this information to characterize the extent to which each group shows rising, stable, or decreasing psychological distress over time, determining whether differences are statistically significant based on 95% CI.

To demonstrate that the analytic subsamples of NHIS respondents before and after each dramatic societal event are comparable across observable demographic characteristics, we provide in SI Appendix, Table S2 descriptive statistics regarding age (in years), gender (proportion female), ethnicity (proportion Latino), national origin (Mexican or Central American), income (proportion with annual household income less than $50,000), length of residence in the United States (in years), region of residence (proportion in Northeast, North Central/Midwest, South, or West), marital status (proportion married or cohabiting), parental status (proportion of parents; proportion with children under 18; proportion with children under 5), and interview language (proportion interviewed in a language other than English). We observe no substantive differences in these characteristics before and after each event within the categories of Latino US-born citizens, naturalized citizens, and noncitizens.

One potential concern is measurement variance. If the different racial/ethnic and citizenship groups we analyze understand and respond to the K6 questionnaire differently, or if their understandings and responses change across the study period, then our analyses may not reflect reliable comparisons across groups. SI Appendix, Table S2 underscores that this is not the case for our sample, and we are further reassured by the secondary literature, which shows that measurement variance is not a concern for the NHIS’ K6 scale (84). We validate this conclusion using multigroup confirmatory factor analysis in SI Appendix, Tables S9–S11 and highlight that the K6 scale as well as our measures of anxiety and depression are invariant across ethnoracial group, citizenship status, region of birth, interview language, and years since immigration.

We use a local linear RDD to examine the impact of the six selected dramatic societal events, each of which we treat as a separate discontinuity. RDD is a causal inference approach in which the “treatment” is a cutoff point on a continuous outcome. We rely on the survey completion date as a running variable, with the date of each societal event representing a cutoff point. Given that policy is exogenous, the cutoff is arbitrary and non-manipulable. Regression discontinuity methods rely on the continuity assumption that individuals who completed their survey just before the event are comparable to individuals who completed their survey just after; in other words, that assignment is ignorable within a narrow interval around the cutoff. In essence, we use the people who completed their interview just before the event as a comparison group to people who completed their interview just after.

Based on the visualization of descriptive trends and the mean-squared error-optimal bandwidth selection process proposed by Calonico and colleagues (51, 52), we selected a 30-d bandwidth on either side of each event to balance between a sufficiently narrow window to capture event effects and a sufficiently wide window to include a reasonable sample size (85).

Our approach is like a regression discontinuity in time approach, except we assume the exogeneity of the interview date with respect to the policy event threshold (85). While the NHIS does allow respondents to schedule an interview, the survey is designed so that respondents will complete the survey when a fieldworker comes to their home (86). We assume a sharp design in which the treatment effect is not time-varying; since we cannot fully determine whether a respondent was made aware of the event at time zero, we measure the intention-to-treat effect.

Based on visual assessments of the dependent variables, we model trends as linear. We use OLS regression for the three continuous measures of psychological distress (i.e., K6, anxiety, depression) and a LPM for the binary measure (i.e., moderate-to-severe distress). We use LPM for interpretability; results from logit models are similar and available in the SI Appendix, Table S3. The equation can be represented as:

$$y_i={\beta }_0+{\beta }_1{\mathit{date}}_i+{\beta }_2{\mathit{cutoff}}_i+{\epsilon }_i,$$

in which date is the date of survey completion and cutoff is a binary variable indicating survey completion on or after the event date for individual i. We present model results as figures, with several concessions to NCHS review. Although it is best practice to include binned daily averages of the outcome in regression discontinuity figures (87), NCHS disclosure risks foreclosed this possibility. Our figures are constructed representations of the RDD models. Additionally, while we use the exact date of survey completion in the model, the results presented in the figures are labeled at the month level at the request of the NCHS.

Beyond measurement variance, another possible threat to our RDD regards the demographic composition of the subset of NHIS respondents analyzed for each dramatic societal event. If the demographic composition of the different citizenship subgroups of Latino respondents changes across the study period, whether in terms of their countries of origin, their years of residence in the United States, or the language in which they completed the survey, then our results may mischaracterize the effect of the dramatic societal events on each groupings’ psychological distress. In SI Appendix, Table S2, we show that the demographic composition of respondents along the categories of country of origin, years lived in the United States, and survey interview language is comparable within the 30-d windows prior to and following each event. We also show in SI Appendix, Table S13 that our results are robust to statistical adjustment for length of US residence and survey language.

As a falsification test, we identify non-Hispanic Black and White respondents who are US born as a single comparison group. The purpose of the falsification test is to examine whether the selected dramatic societal events and/or the country’s everyday institutional and social environments of deportation threat are, on average, associated with this group’s psychological distress. This falsification test is common in studies of immigration enforcement’s effects on anxiety- or stress-related outcomes, as there is little theoretical reason to expect that changes in the threat of deportation should impact non-Hispanic Black and White people who are US-born citizens (88–90). We have similar expectations in the current study. Most of the selected dramatic societal events concern DACA and DAPA, which are responses to immigration authorities’ disproportionate targeting of Latino noncitizens, their families, and their communities. Accordingly, the announcement and rescission of those initiatives should affect Latinos across citizenship statuses in a way that they should not impact non-Hispanic Black and White US-born citizens, on average. With respect to the 2016 US presidential election, we posit that it should affect Latinos across citizenship statuses but not impact non-Hispanic Black and White US-born citizens, on average, given the vitriolic campaign rhetoric targeting Latino noncitizens, their families, and their communities.^¶ For each dramatic societal event, we replicated our RDD and observe no significant discontinuities in K6 scores or rates of moderate or severe psychological distress for non-Hispanic Black and White US-born respondents overall. In SI Appendix, Fig. S4, we do not observe significant discontinuities in either measure for non-Hispanic White and Black US-born citizens separately either, except for a marginal decrease (P < 0.10) in non-Hispanic Black US-born citizens’ K6 scores following the 2016 presidential election; this decrease is nonetheless oppositely signed relative to our observed results for Latino noncitizens.

Our analyses of the country’s everyday institutional and social environments of deportation threat also rely on an OLS regression approach for the three continuous measures of psychological distress and a LPM for the binary measure. We include two key independent variables. To measure the institutional environment, we use monthly counts of the number of detainer requests ICE issued between January 2011 and December 2018 from Syracuse University’s TRAC (53). To measure the social environment, we rely on publicly available Google Trends data regarding searches for the Immigration topic every month between 2011 and 2018 (54). We standardize both variables, whose correlation is −0.13.

We select detainer requests as our measure of the institutional environment of deportation threat for several reasons. One is that it is a measure of interior enforcement. As a household survey, the NHIS samples Latinos settled in the country; measures that capture interior enforcement, therefore, better reflect their everyday experiences. Another is that detainers are an indication of both everyday and immigration policing (91), each of which bears on Latino noncitizens’ perceived risk of deportation. Subnational police must first arrest and fingerprint someone before the DHS can issue a detainer request. Not all who are arrested are processed for deportation, but increased arrests suggest that police officers are detaining people whom ICE has an interest in deporting. Police arrests accounted for 55 percent of all deportations stemming from interior immigration enforcement activities in 2014 and 2015 (31, 32). Detainers reflect localized enforcement practices nationwide, as an established literature shows (31, 32).

Our measure of the social environment of deportation threat from Google Trends reflects the national salience of immigration enforcement to the US public. Unlike interview- or survey-based measures that ask participants to report their awareness or fears of deportation, Google searches are unlikely to suffer from social desirability bias because they are done in private (92, 93). They do not measure attitudes, but they are at minimum a measure of interest in or behavior around deportation (55, 56). Several studies have suggested the same (88, 94). As a measure of the national salience of the Immigration topic, we prefer the Google Trends measure to other possible measures such as the volume of news articles on the topic. The fact of a news article’s existence says nothing of its public consumption. By contrast, the Google Trends measure is constructed based on users’ search behaviors and, therefore, should better reflect how salient the Immigration topic is to a public that is seeking out related information. Google Trends makes available an alternate measure of the relative volume of Google News searches for the Immigration topic across the study period; this measure yielded substantively similar results and is presented in SI Appendix, Fig. S8.

Google Trends provides us with two additional points of information that help substantiate our argument that it reflects the national salience of deportation to the US public. One is the list of the top 20 queries related to the Immigration topic, which reflects the search terms Google users put into the search engine. These related queries include all the dramatic societal events we considered, as well as several other aspects of the broader national conversation around deportation during the study period (e.g., “immigrant caravan” and “immigrant children separated from parents”).^# Another point of information is the list of the top ten related topics that users who searched for the Immigration topic also searched for. As above, these capture the dramatic societal events we considered, as well as several other aspects of the country’s everyday social environment of deportation threat during the study period.^||

In addition to the two key independent variables, our regression analyses of the public-access NHIS data include a range of demographic controls and fixed effects. We adjust for gender (1 = female), age (in years), employment status (1 = working in the last week), income (1 = household income $50,000 or greater), parental status (1 = is a parent), marital status (1 = married or cohabiting), education (less than high school, high school, some college, or bachelor’s), region of residence in the United States (Northeast, North Central/Midwest, South, or West), and interview language (1 = interview language other than English). We also include fixed effects for the month and the year of the NHIS interview. The model can be represented as:

$$y_{\mathit{ijk}}={\beta }_0+{\beta }_1{\mathit{detain}}_{\mathit{jk}}+{\beta }_2{\mathit{google}}_{\mathit{jk}}+{\mathit{X}}_{\mathit{i}\mathit{j}\mathit{k}}{\mathit{\beta }}_{\mathit{i}\mathit{j}\mathit{k}}+{\alpha }_j+{\gamma }_k+{\epsilon }_{\mathit{ijk}},$$

in which detain is the standardized measure of detainers and google is the standardized measure of Google searches in month j and year k, X is a vector of demographic controls for each individual i, ${\alpha }_j$ represents month dummy variables, and ${\gamma }_k$ represents year dummy variables. In SI Appendix, Table S6, we also considered an interaction term for the role of poverty status using a binary indicator for whether the respondent’s household falls below the poverty threshold (generated from POORYN, an IPUMS-NHIS indicator that accounts for household income, family size, and the number of children in the household under age 18), given prior research suggesting that socioeconomic status can pattern perceptions of deportation threat among Latino naturalized citizens and noncitizens (15, 16).

As with the RDD, we replicate our modeling strategy for non-Hispanic Black and White US-born citizen respondents. We find that the association of the standardized detainers and Google Trends measures are oppositely signed relative to those for the Latino citizenship subgroups and/or statistically insignificant. In SI Appendix, Table S5, we show that this result is robust to separating non-Hispanic Black and White US-born citizen respondents into two distinct groupings.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

A replication package containing all code used in this analysis is available through GitHub (https://github.com/aljohnson-soc/deportation-threat) (95). The public-access NHIS data are available for download through IPUMS-NHIS (https://nhis.ipums.org/nhis/) (34). The restricted-access NHIS data are available with Special Sworn Status and via application to the NCHS (https://www.researchdatagov.org/product/10480) (35). The detainers data are available with application to TRAC (https://trac.syr.edu/fellows/) (53). The Google Trends data are available for download through Google Trends (https://trends.google.com/) (54).