Introduction

Nearly 3.5 million children of Mexican descent in the United States live with an unauthorized immigrant parent, even though the vast majority of these children (88 %) are US-born citizens [1]. The severe disadvantages these families face, (e.g., poverty, deportation fears, and social marginalization) heighten their risk of food insecurity [2–6]. Food insecurity, defined as not having access to nutritionally adequate food due to financial constraints, is a severe health risk that increases children’s susceptibility to infections and chronic diseases, slows physical growth, and reduces health into adulthood [7, 8]. Indeed, evidence suggests that higher food insecurity alone explains much of the poorer overall child health observed in immigrant families [9].

In recent years, state and local authorities have become more involved in enforcing immigration laws, which could potentially increase the food insecurity risk of unauthorized Mexican immigrant families. Federal immigration reform adopted in 1996 set the stage for today’s localized immigration enforcement efforts by delegating federal immigration powers to state and local governments for the first time through the 287(g) program. 287(g) allowed local law enforcement to enforce federal immigration laws during routine policing activities and was intended to target criminal offenders. However, some localities have used it as a universal means to deport unauthorized immigrants by raiding homes and businesses, setting-up driver’s license check-points, and conducting traffic stops for minor offenses [10, 11]. These efforts have largely targeted Hispanics and Hispanic communities [11].

Responding to criticisms, Immigration and Customs Enforcement officially ended 287(g) in 2012 but remains committed to working with local law enforcement. 287(g)’s successor program, Secure Communities (piloted in 2008) eliminated immigration enforcement during routine policing activities and focused on data sharing of arrestees. The Secure Communities program, however, continued to disproportionately target Hispanics and to be abused/misused by local law enforcement [12]. As a consequence, in 2014 the Secure Communities program was replaced by the even narrower Priority Enforcement Program, which focuses on data sharing of convicted criminals. Despite these narrower program efforts, concerns about localized abuse/misuse persist [10], making it essential to understand the consequences of such policies starting with the foundational policy, 287(g).

Local immigration enforcement can influence household food insecurity three main ways. First, deportations from 287(g) and other localized immigration enforcement efforts increase the economic disadvantage of family members left behind [2, 3, 10, 13–15]. An estimated 90,000 parents of US-born citizen children are deported each year [10]. Typically, fathers are deported; the remaining single-mother headed households now face lost income, legal fees, and family reunification costs [14]. As a consequence, these single mothers struggle to provide the most basic necessities, including food, for their children [3].

Second, deportations also increase fear and mistrust among immigrants, which may reduce immigrant use of social services that protect against food insecurity. Research indicates that unauthorized immigrants’ heightened sense of “deportability” makes them fearful of the community and distrustful of public agencies [3, 13, 15–18]. As a result, although US-born children in unauthorized immigrant families are eligible for food stamps and other social services, they are often confused about eligibility and fearful that applying for benefits could cause deportation [19], particularly in areas with heightened immigration enforcement [14, 18, 20, 21].

Third, fears of deportation, family separation, and police harassment also decrease unauthorized immigrants’ mobility and increase their social isolation and emotional distress [18, 22, 23]—all of which have implications for food insecurity. Qualitative evidence, indicates that to avoid police immigrants often refrain from driving, which could alter food consumption patterns as families rely more on fast-food and high-priced, walking-distance grocery stores [10, 23]. Additionally, constant worry about family separation leads to social isolation and strains social ties as unauthorized immigrants avoid public places, including churches and schools, which may provide food supports [13, 24]. Social isolation and worry about deportation also exacerbate parental distress, depression and anxiety [18, 25]—known risk factors for food insecurity [7, 8].

The worry generated by 287(g) might not be limited to unauthorized Mexican immigrants but may have broader spillover effects in the Hispanic community, which feels targeted by anti-immigrant policies [26]. Trust, reciprocity, social networks are known protectors against food insecurity [27]. If these protective factors deteriorate in Hispanic communities [13, 18], and/or if Hispanic parents face greater levels of depression due to perceived policy discrimination [28], even Hispanic citizen families may experience an increase in food insecurity.

Growing research confirms that deportations and greater local immigration enforcement lowers social service use and increases economic hardship [15, 18, 21, 29, 30], but most studies are based on qualitative data and small-scale surveys, which provide rich contextual information but are limited with respect to representativeness. Moreover, these studies have not demonstrated that 287(g) increases food insecurity. We address this gap and answer calls for research on the health and well-being impact of local level immigration policies [11, 23] by examining how 287(g) impacts food insecurity among Mexican non-citizen households—the group with the highest prevalence of undocumented immigrants [1] and levels of deportation [10]. We also examine a sub-sample of low-income households, which are most at risk of food insecurity [7–9]. Finally, we examine Hispanic citizen households to see if a spillover effect exists. In our analysis, we focus on county- and city-level 287(g) programs, which account for over 80 % of programs and have garnered the most concern about abuse/misuse.

Methods

We analyze data from the Current Population Survey Food Supplemental Survey (CPS-FSS). The CPS collects monthly demographic and employment information from about 60,000 housing units across the US. The CPS-FSS is a supplementary questionnaire consistently administered in December that focuses on household food consumption. The CPS-FSS has a nationally representative sample spanning multiple years, which makes it possible to assess heterogeneous effects by citizenship and race/ethnicity, and to conduct analysis spanning pre-post-287(g) adoption.

The CPS-FSS also identifies the core-based statistical area (CBSA) of residence for individuals living in socioeconomically tied urban centers with at least 10,000 people. These CBSAs allow us to identify geographic residence for households living in large metro-areas and cover nearly 90 % of all local 287(g) programs. Although 287(g) programs are adopted at the county- and city-level, the broader metro-area is an appropriate geographic level of assessment. Even in cases where a 287(g) program covers only part of a metro-area, the entire metro-area is likely affected because individuals often cross city/county lines on a daily basis to get to work/school. Moreover, evidence suggests individuals mistakenly assume their own local law enforcement agency has adopted 287(g) when neighboring agencies have the program [15]. Because data are public-use this project received exempt IRB status.

Sample

We pool 2004–2009 CPS-FSS data to obtain multiple pre-post-287(g) observations. We limit pre-observations to 2004 to obtain a consistent CBSA classification and post observations to 2009 because 287(g) was weakened and subjected to more federal oversite in 2009 [16] and to avoid confounding effects of the Secure Communities program [15]. Our combined sample is 58,353 households, which was reduced due to missing data on food insecurity (N = 155) or an independent variable (N = 1029). Our primary sample of interest is Mexican non-citizen households with children ages 0–17 (N = 3307). Following the convention in literature [15, 31–33], this sample serves as a proxy for unauthorized Mexican households, which are unidentifiable in most data. Mexican non-citizen households include both mixed citizen and all non-citizen households, both of which are vulnerable to 287(g). We use Hispanic citizen households (N = 4710) to assess spillover effects and include non-Hispanic white (N = 40,427) and black citizen (N = 7905) households as a sensitivity check. We examine the full sample for each group and a sub-sample of low-income households 185 % below the federal poverty line.

Measures

Dependent Variable

The CPS-FSS includes a binary classification of food insecurity status where a “1” is assigned for food insecure if three or more items on the US Department of Agriculture’s 18-question survey were answered in the affirmative [8]. The food insecurity measure is based on experiences over the past 12 months.

Independent Variable

Our main independent variable is a binary policy indicator that represents whether individuals reside in a metro-area with a 287(g) program in year t − 1. We assess the lagged program effect because food insecurity is a 12-month retrospective measure and because it takes time to implement a 287(g) program [16].

Control Variables

We control for household characteristics associated with food insecurity: household size, age of oldest child, and head of household’s education, marital status, age, and years in the US [7, 8]. To account for metro-area differences, we include metro-area fixed effects that control for time-invariant characteristics unique to each metro-area. We also include time-varying metro-area economic (unemployment and poverty rates) and demographic (percent foreign-born) conditions. Lastly, we include year fixed effects to account for national trends.

Statistical Analysis

We use linear probability models (LPM) and a difference-in-difference type strategy to identify the influence of 287(g) on household food insecurity. LPMs with robust standard errors are analogous to the logit model but are more computationally efficient for fixed effects [34, 35]. To adjust for heterogeneity, all analyses are weighted and clustered on metro-area-year of residence.

The difference-in-difference method has been extensively used to identify the effects of local enforcement policies [15, 31–33, 36] and compares food insecurity risk before and after 287(g) adoption for a treatment and comparison group. Table 1 lists metro-areas with 287(g) identifiable in the CPS; these make-up the treatment group. Metro-areas without 287(g) are the comparison group. About 10 % of metro-areas identifiable in CPS (N = 278) have adopted 287(g). Because 287(g) was adopted at different points in time, there is not a single pre-post indicator. Instead, year fixed effects capture time differences, and metro-area fixed effects capture metro-area differences.

Table 1 List of metro-areas in the current population survey that adopted a 287(g) program between 2004 and 2009
Full size table

The difference-in-difference strategy requires that the adoption of 287(g) be exogenous to food insecurity. Local law enforcement agencies, however, chose to apply for 287(g). First, the adoption of 287(g) may have been associated with growth in the foreign-born/Hispanic population [37, 38], which may also be related to food insecurity [39]. Thus, we run a sub-analysis that includes only metro-areas with similar growth pre-adoption, i.e., in the top 50th percentile of foreign-born growth between 1990 and 2000. Second, unobserved political and social factors may have influenced 287(g) adoption [38] and food insecurity. Thus, we run an additional analysis using only metro-areas that applied for 287(g) but were denied/withdrew [40]. We assume these metro-areas are most similar to those with 287(g).

The difference-in-difference model also assumes that in the absence of 287(g) the groups experience similar food insecurity trends. The Great Recession (2007–2009) might threaten this assumption because it coincided with the timing of many 287(g) programs. To account for this possibility, we run the analysis for non-Hispanic white and black citizens, who were similarly affected by the recession [41] but not 287(g).

Finally, because law enforcement agencies differed in their vigor of 287(g) enforcement [16, 33], we run analysis using proxy indicators of enforcement intensity. We collected annual information on the number of individuals identified for removal under 287(g) for each law enforcement agency [42] and classified 287(g) programs as low (i.e., had no removals) and high (i.e., had removals) enforcement intensity.

Results

Table 2 provides summary statistics for Mexican non-citizen households with children in metro-areas that adopted and did not adopt 287(g). For both the full and low income samples metro-areas with and without 287(g) have similar characteristics. Length of residence in the US and household marital status are not statistically different and differences in the average household size, education level, and age of household head, though statistically significant, are minimal in size. As expected, we do find that metro-areas adopting 287(g) have a higher concentration of foreign-born residents (21 vs. 16 %). Overall, these similarities provide support for a difference-in-difference design but suggest robustness checks are necessary.

Table 2 Weighted sample summary statistics of Mexican foreign-born non-citizen households with children
Full size table

Multivariate Analysis

Table 3 presents the linear probability model results. We present an unadjusted model that includes a 287(g) dummy variable and only metro-area and year fixed effects, as well as an adjusted model with all controls. We present the results for the full and low-income samples and for different ethnic/racial groups. Panel I presents results using all metro-areas without 287(g) as the comparison group, Panel II restricts the comparison group to metro-areas with high foreign-born growth, and Panel III restricts the comparison group to metro-areas that requested but did not adopt 287(g).

Table 3 Linear probability model of the association between 287(g) and household food insecurity risk for households with children
Full size table

The results suggest two trends. First, results provide strong support that 287(g) is associated with food insecurity among Mexican non-citizen households with children. Specifically, Panel I shows that in the full sample (Part A) the adoption of 287(g) is statistically significant and associated with a 9.9 percentage point increase in food insecurity risk in the unadjusted model and a 10.9 percentage point increase in the adjusted model. These results remain relatively unchanged when we focus on the more vulnerable and economically comparable low-income sub-sample (Part B). Second, we find little evidence that 287(g) has broader spillover effects on the food insecurity risk of Hispanic citizen households. Examining Panel I and the full sample (Part A) we find a small, positive (1.9 percentage points) association in the unadjusted model, but this result is not statistically significant in the adjusted model nor in the low-income sub-sample in Part B.

Sensitivity Analysis

Table 3 also reports results from the sensitivity analysis. To ensure our results do not reflect alternative shocks, e.g., the Great Recession, we run the analysis on two groups not targeted by 287(g): non-Hispanic white and black citizens. We find no association for the non-Hispanic white sample and negative but weak association for the non-Hispanic black sample. If alternative economic shocks were driving the trends in food insecurity rates rather than the 287(g) program, the non-Hispanic white and black citizens should have been similarly affected.

To address policy endogeneity concerns we restrict the comparison group to metro-areas with high foreign-born growth (Panel II) and to metro-areas that have applied for 287(g) but were denied/withdrew (Panel III). The general robustness of the results in Panels II and III suggests that policy endogeneity is not driving the results. Although the sample sizes for Mexican non-citizen households are notably smaller in Panels II and III, the coefficients in the adjusted models for the full and low-income samples remain robust (a near 10 percentage point increase) and significant or marginally significant.

Table 4 presents results for 287(g) program enforcement intensity. We find that 287(g) programs, no matter their enforcement intensity, are associated with increased food insecurity risk. For the low-income sample, however, this increase in food insecurity risk is greater when enforcement is more intense. In the adjusted model, 287(g) programs with low and high enforcement intensity are associated with a respective 8.8 and 13.1 percentage point increase in food insecurity risk. The difference between these two estimates is statistically significant [F(2575) = 4.87; p < 0.01].

Table 4 Linear probability model of the association between 287(g) program enforcement intensity and household food insecurity risk for Mexican non-citizen households with children
Full size table

Discussion

Our research provides the first national-level evidence that local immigration enforcement laws negatively influence the health and well-being of immigrant families, specifically Mexican non-citizen families. We examine the consequence of 287(g), the foundational local immigration enforcement program, for one critical risk factor of child health and development—food insecurity [7–9]. For Mexican children of immigrants who are already at increased risk for poor health [5, 6], household food insecurity is likely to have a strong influence on their future social and economic integration.

We found that 287(g) was positively associated with food insecurity risk among Mexican non-citizen households with children. Mexican non-citizen households residing in a metro-area that adopted 287(g) were 10 percentage points more likely to experience food insecurity than their peers in metro-areas without 287(g). This is a substantial increase for a population that already experiences high levels of food insecurity (approximately 30 % in our sample). The associations were similar when we focused on the more vulnerable low-income households. Additionally, we found that food insecurity risk was greatest in metro-areas that had used the program to remove unauthorized immigrants (i.e., high program enforcement intensity). These results support findings from smaller-scale survey and qualitative research, which indicate deportations and local immigration enforcement heighten a community’s sense of deportability and severely disrupt children’s home life leaving them vulnerable to food scarcity and other preventable health risks [3, 10, 13, 15, 18]. That said, we found no evidence of negative spillover effects on the food insecurity of Hispanic citizen households with children. Even though Hispanics, no matter their citizenship status, likely experience negative effects of 287(g) and similar programs [10, 13, 15], these effects do not appear to extend to food insecurity.

Limitations

A causal interpretation or our results should be done with care. Our results may be partially capturing negative effects of the Great Recession and policy endogeneity remains a concern. We cannot rule out the possibility that immigrants moving to or out of metro-areas with and without 287(g) are systematically different, which could bias our results. It is also possible that we underestimated the association between 287(g) and food insecurity because of data limitations in identifying documentation status. Our results represent the overall influence on Mexican non-citizen households, both legal permanent residents and unauthorized immigrants.

Conclusion

Though 287(g) has officially ended, federal efforts to partner with local law enforcement to enforce immigration laws persist. Our results suggest that these continued efforts may have unintended consequences on the health and well-being of immigrant families with children. The narrower focus of these federal-local partnerships is an improvement and will likely reduce unintended health consequences for immigrant families with children. However, important concerns remain. First, the newer federal-local partnerships have been more widely adopted than 287(g); for example, nearly all US counties participated in Secure Communities. Thus, more immigrant families are susceptible to the unintended consequences of these programs. Second, the persistence of reported local abuses despite continual shifts in policy suggests that challenges do not lie in a particular policy, but are rather inherent in local law enforcement’s involvement in immigration enforcement. As such, the newer, narrower policies are unlikely to avoid ill effects on immigrant families. Lastly, even in newer policies, children remain vulnerable. Almost 40 % of individuals apprehended under the Secure Communities program were the parents of US citizen children [12]. As long as any immigration policy is in place that seeks to apprehend and deport adults, support systems such as access to food stamps, health care, and mental health services [14] need to be put in place to protect the health and well-being of children and their remaining caregivers who are left behind.