Family Separation and Reunification Under President Trump's Zero-Tolerance Policy

Abstract

The past decades have been characterized by sharp increases in the number of families, mainly from Central America's Northern Triangle, apprehended by US Border Patrol. In an effort to stem those flows, the Trump administration implemented a zero-tolerance policy (ZTP) aimed at criminally prosecuting all adults crossing the border without authorization, regardless of whether they traveled with children or sought asylum upon entry. Thousands of children were separated from their parents, reclassified as “unaccompanied,” and referred to the custody of the Office of Refugee Resettlement (ORR). Yet, to date, there has not been a careful evaluation of the impacts of the policy. We examine how ZTP affected the volume of unaccompanied minors, their time in ORR custody, and their likelihood of family reunification. We show that ZTP boosted the ranks of unaccompanied children through family separations by 48 percent, lowered their discharge rate from ORR's custody by 38 percent, and reduced their odds of family reunification by 49 percent. Given the growing number of families from the Northern Triangle seeking asylum in the United States, the documented mental health problems of separated children, and the rotating nature of immigration policies based on the administration in place, understanding the implications of policies like ZTP is imperative.

Keywords

family separation family reunification zero tolerance policy unaccompanied minors.

I have put in place a ‘zero-tolerance’ policy for illegal entry on our Southwest border. If you cross this border unlawfully, then we will prosecute you. It's that simple. If you are smuggling a child, then we will prosecute you, and that child will be separated from you as required by law.¹ (Jeff Sessions, May 7, 2018)

Introduction

The number of unaccompanied minors apprehended by immigration authorities along the US-Mexico border rose by an astonishing 300 percent over the past decade, from 17,909 in 2010 to a peak of 70,418 children in 2019 (CBP 2021b). That figure has continued to rise, reaching 144,834 minors during FY2021 (CBP 2021a). Most studies on the root causes of such an increase point to violence and poverty in the Northern Triangle countries of El Salvador, Guatemala, and Honduras as the primary drivers (UNICEF-UNHCR 2020; Clemens 2021). Others have also underscored the role played by policy changes (e.g., Amuedo-Dorantes and Puttitanun 2016). In this study, we explore the impact of a recent and highly controversial policy from the Trump administration — namely, the zero-tolerance policy (henceforth, ZTP) — on boosting the ranks of unaccompanied minors through the reclassification of children after family separations, as well as on the children's family reunification prospects. Despite the media attention drawn by ZTP, to date, there has not been a careful evaluation of this policy's impacts.

On April 6, 2018, then-US Attorney General Jeff Sessions announced a sweeping “zero-tolerance” policy aimed at curtailing the growing flows of families reaching the US-Mexico border, most of whom had been turning themselves into Border Patrol officers seeking asylum. Until then, unauthorized border crossers who were not an enforcement priority, as in the case of most individuals in family units, were usually placed in civil removal proceedings for unauthorized presence but not criminally prosecuted (CRS 2021). This practice changed with the implementation of ZTP, as it required all adults entering without authorization to be criminally prosecuted, regardless of whether they traveled with children or sought asylum upon entry.

As adults were referred to the US Department of Justice for prosecution and placed in federal criminal facilities, children were separated from their families to comply with the Flores Settlement agreement requiring minors to be held in the “least restrictive setting.” Furthermore, following the 2008 Trafficking Victims Protection Reauthorization Act (TVPRA), minors originating from contiguous countries, such as Mexico, were returned home, whereas those from non-contiguous countries were reclassified as “unaccompanied” and placed under the Office of Refugee Resettlement (ORR) custody (Ryo and Humphrey 2021), barring “exceptional circumstances.” Due to the differential treatment of minors regulated in the Flores Settlement agreement and the 2008 TVPRA, ZTP likely raised the number of unaccompanied minors in the custody of US Border Patrol agents through the separation of thousands of families apprehended at the border, most of which were coming from the Northern Triangle (ACLU 2018).

Even though President Trump issued an executive order on June 20, 2018, ending the policy amid public outrage and political pressure (Department of Justice 2021), more than 2,000 children were still waiting to be reunified with their families three years later (Interagency Task Force on the Reunification of Families 2021). To date, despite the damage caused, we have a limited understanding of the extent to which ZTP boosted the ranks of unaccompanied minors through family separations, lengthened the time children were held under ORR custody, and harmed family reunification prospects due to lack of planning and proper communication.

We address that gap in the literature using two datasets gathered through Freedom of Information Act (FOIA) requests from the Department of Homeland Security. First, using Border Patrol data on monthly apprehensions of unaccompanied minors along the southwest border by country of origin, we explore how ZTP increased the number of unaccompanied children through family separations. To that end, we compare pre- versus post-policy changes in the volume of unaccompanied minors from the Northern Triangle, who must remain in ORR custody once separated from their families and reclassified as unaccompanied minors, to pre- versus post-policy changes in the volume of unaccompanied minors from Mexico, who are usually returned. We find that, through the separation of families involved in the policy's implementation, ZTP raised the number of children classified as unaccompanied minors by 48 percent.

Subsequently, we use a second dataset on unaccompanied minors referred to ORR from October 2015 through October 2019 to explore how ZTP lengthened children's time under the agency's custody and harmed their family reunification prospects. Estimates from survival and competing risks models reveal the very damaging impact of ZTP on both outcomes. We find that children's discharge rate decreased by 38 percent, lengthening their duration under ORR's custody. Furthermore, their odds of family reunification dropped by 49 percent and, even one year after being placed under ORR's watch, family reunification remained almost 20 percent less likely than before ZTP was implemented. According to internal government reviews, the Department of Homeland Security failed to coordinate with other departments, including the Department of Health and Human Services, on the implementation of ZTP and to provide notice or guidance for the separation of children (Department of Justice 2021). The lack of coordination did not allow ORR to prepare for the increase in the volume of unaccompanied minors, left the department “unable to provide prompt and appropriate care for separated children,” and presented operational challenges for facilities that housed immigrant minors (Department of Health and Human Services 2020). It also caused delays in family reunifications as there were no policies to facilitate communication between children in ORR custody and their parents. In some cases, this resulted in delays of up to two months before separated minors and their parents could establish contact (Department of Justice 2021).

Overall, the study provides insight into the impact of the Trump administration's ZTP on the volume of unaccompanied minors, the duration of their custodial arrangements, and their odds of family reunification. Given the increasing number of families from the Northern Triangle countries traveling with young children and seeking asylum in the United States, along with the widely documented greater risk of depression, post-traumatic stress, and other mental health problems among children in institutional settings (e.g., Linton, Griffin and Shapiro 2017; Carey 2018), understanding the implications of this type of policy approach to immigration enforcement is well-warranted.

Background and Legal Framework

As recently as FY2011, children traveling alone or as part of a family unit represented only 7 percent of migrants apprehended while crossing the southwest border. Their share increased almost four-fold to 25 percent of all apprehensions during FY2016, and to 38 percent by 2019.² Most of these minors migrated from Central America's Northern Triangle, escaping widespread poverty and violence exacerbated by political instability, demographic and environmental factors at home (Capps et al. 2019). At the same time, a strong labor market and the possibility of family reunification in the United States provided further incentives for children traveling North. US immigration authorities, who had largely succeeded in deterring adult migrants from Mexico through expedited removals and the criminal prosecution of reentries, could not employ the same strategies in response to the surging flow of Central American families or minors.³

To understand how ZTP impacted the volume of apprehended unaccompanied minors and their time in custody and family reunification prospects, it is critical to consider the legal framework in which the policy was adopted. While ZTP mandated the criminal prosecution of adult migrants, three crucial pieces of legislation regulating the processing, treatment, and release of migrant children created the legal context that ultimately determined the outcomes of thousands of children crossing in the company of their adult relatives. Figure 1 includes a diagrammatic description of that legal framework, which included the Flores settlement agreement reached in 1997, the Homeland Security Act of 2002, and the 2008 TVPRA (Ryo and Humphrey 2021).⁴ Overall, the legal framework differentiated unauthorized migrant children based on two dimensions: (1) whether they were apprehended while traveling alone or with family, and (2) whether they originated from a contiguous or a non-contiguous country. These two dimensions determined how a minor was processed. To understand why, it is critical to know what these regulations stipulate.

Figure 1.

Zero-tolerance policy legal framework.

The Flores settlement agreement emanated from a lawsuit filed in 1985 by migrant children against the government to challenge the procedures involved in their detention, treatment, and release. The parties reached an agreement in 1997, stating that children under government custody should be held in the “least restrictive setting.” This meant that — until the adoption of ZTP — apprehended family units were usually released as they waited for their immigration court hearings, given that children could not be kept in detention facilities. With the activation of ZTP, however, Border Patrol started referring all migrant adults to the Department of Justice for criminal prosecution regardless of whether they were apprehended traveling with children or seeking asylum. As adults were placed in detention facilities, authorities separated family units and reclassified children as “unaccompanied” minors under the argument that the Flores agreement prevented them from keeping families together.⁵

The 2002 Homeland Security Act dissolved the Immigration and Naturalization Service (INS) under the Department of Justice and created the Department of Homeland Security. The new department included three agencies responsible for enforcing immigration law — US Citizenship and Immigration Services, Immigration Customs Enforcement, and Customs and Border Protection. The law also transferred authority over the processing and care of unaccompanied minors from the old INS to the ORR, under the US Department of Health and Human Services, to ensure that the least restrictive setting would be observed, as regulated in the Flores settlement agreement. In addition, it mandated the release of children to parents, family, or other sponsors “without unnecessary delay.”

At the same time, the 2008 TVPRA regulated the processing of unaccompanied minors. The law established a critical distinction based on whether unaccompanied migrant children originated from contiguous countries (i.e., Canada and Mexico) or non-contiguous countries. Children in the first group were to be repatriated if there were no grounds for asylum; otherwise, they were transferred to the custody of ORR. Alternatively, unaccompanied minors from non-contiguous countries, including the Northern Triangle, were placed under ORR custody within 72 hours, barring “exceptional circumstances.”

Amid this legal context, and as flows of single adults and family units seeking asylum increased along the US-Mexico border, the Trump administration announced its ZTP in April of 2018. The policy aimed to criminally prosecute all adults entering without authorization, including those seeking asylum upon entry, regardless of whether they were traveling with children.⁶ While the activation of ZTP did not modify ORR's responsibility over unaccompanied minors, the differential treatment of children by country of origin and by whether they were apprehended while traveling with or without relatives, coupled with growing backlogs in immigration courts due to the criminalization of unauthorized entry into the United States,⁷ led to an ever-expanding number of unaccompanied minors being held under ORR custody.

The situation was exacerbated by contextual factors, including the lack of coordination between different departments involved in the separation and processing of families, as well as ORR's failure to plan for and build the capacity to manage the surge in reclassified unaccompanied minors (Department of Health and Human Services 2020). Figure 2 presents a diagram depicting how these contextual factors affected the treatment of minors under ORR custody during ZTP, which led to extended family separations while the cases were resolved. This resulted in some families back home assuming their children were permanently staying in the United States (Amuedo-Dorantes and Puttitanun 2016). When asylum cases were rejected, parents were commonly deported without their children or even knowledge of their whereabouts (Jones, Obser and Podkul 2017). This treatment of families caused public outrage. On June 20, 2018, President Trump issued an executive order ending the policy; however, thousands of children remained separated from their families three years later (Interagency Task Force on the Reunification of Families 2021). In what follows, we assess how ZTP impacted the volume of unaccompanied minors, the duration of their custodial arrangements, and their family reunification odds.

Figure 2.

Zero-tolerance policy contextual framework.

Data and Descriptive Statistics

Data on Unaccompanied Minors

We work with data gathered through two FOIA requests on apprehensions of unaccompanied minors by US Border Patrol and on unaccompanied minors referred to ORR. The first dataset includes aggregate information on over 330,000 apprehensions of unaccompanied minors by month, border sector, and country of origin conducted by US Border Patrol agents between October 2013 and December 2019 along the US southern border (see Figure 3). These minors encompass children younger than 18, deemed “inadmissible aliens” under Title 8 of the United States Code Section 1227 (Deportable Aliens) while crossing the US border without legal authorization. The dataset includes children traveling alone and those separated from their parents or legal guardians and reclassified as unaccompanied minors under ZTP.⁸

Figure 3.

US Border Patrol sectors along the southwest border.

In Panel A (Table 1), we present summary statistics aggregated at the (month $\times$ Border Patrol sector $\times$ country of origin) level, while differentiating between a pre-ZTP period, spanning from October 2013 through March 2018, a ZTP period when the policy was in place from April through June 2018; and a post-ZTP period spanning from July 2018 through December 2019. The first mean (116.91) represents the average number of apprehended unaccompanied minors (or unaccompanied alien children, UACs) in those cells before ZTP was activated. Throughout the study period, the number of unaccompanied minors from the Northern Triangle apprehended by Border Patrol represented approximately 80 percent of the flow, increasing by about 40 apprehensions per month, country of origin, and Border Patrol sector during ZTP and continuing thereafter, even after the policy had officially ended, albeit at a more moderate rate. In contrast, the number of migrant children from Mexico apprehended every month in each Border Patrol sector dropped by 15 percent during ZTP to barely rise after the policy's suspension. These statistics support the notion that ZTP impacted minors from the Northern Triangle differently than those from Mexico, possibly driven by the fact that 96 percent of separated families during the policy originated in the Northern Triangle (ACLU 2018).

Table 1.

Descriptive Statistics for Unaccompanied Minors, by Period.

Panel A: Unaccompanied minors apprehended by US Border Patrol
	Pre-ZTP	Active ZTP	Post-ZTP
	10/2013–03/2018	04/2018–06/2018	07/2018–12/2019
Monthly UACs	116.91	141.06	145.25
	(269.67)	(240.22)	(255.08)
Monthly UACs from Northern Triangle	119.77	157.51	160.79
	(302.97)	(273.24)	(290.31)
Monthly UACs from Mexico	108.34	91.74	98.65
	(124.45)	(66.35)	(68.36)
Countries of origin characteristics
Homicide rates	44.539	35.625	33.833
	(24.317)	(11.162)	(9.096)
Employment-to-population ratio (%)	0.594	0.603	0.605
	(0.022)	(0.031)	(0.031)
Human Development Index	0.677	0.684	0.686
	(0.054)	(0.055)	(0.055)
Under-five mortality rate	19.446	17.782	17.392
	(5.136)	(4.600)	(4.487)
Child mortality rate (age 5–14)	3.642	3.518	3.477
	(0.827)	(0.730)	(0.705)
Observations	1,944	108	648

Table 1. (continued)
Panel B: Unaccompanied minors referred to ORR
	Pre-ZTP	Active ZTP	Post-ZTP
	10/2015–03/2018	04/2018–06/2018	07/2018–10/2019
Age	14.599	14.341	14.740
	(3.269)	(3.517)	(3.257)
Female	0.318	0.291	0.331
	(0.466)	(0.454)	(0.470)
Length of time in ORR custody (days)	41.638	64.166	46.797
	(49.112)	(56.200)	(41.137)
UACs from Northern Triangle (%)	0.941	0.937	0.925
	(0.235)	(0.242)	(0.264)
UACs reunified with family (%)	0.948	0.830	0.940
	(0.222)	(0.375)	(0.238)
Number of monthly ORR referrals	4,729.31	5,251.74	6,018.97
	(1,864.03)	(691.21)	(2,233.13)
Observations	111,874	15,467	81,879

Note: Panel A shows summary statistics for the universe of unaccompanied minors apprehended by Border Patrol along the southwest border between October 2013 and December 2019, as well as country-of-origin characteristics included in our empirical model. The data are aggregated at the month $\times$ Border Patrol sector $\times$ country of origin level. Panel B presents descriptive statistics for all unaccompanied minors from Mexico and the Northern Triangle (Salvador, Guatemala, and Honduras) placed in ORR custody between October 2015 and October 2019. The data is at the individual level. Sample means, standard deviations in parentheses.

As noted in the previous section, children from non-contiguous countries and, in exceptional cases, Mexico are referred to ORR for their custody and care. ORR determines their temporary placement and, when applicable, reunites them with family members or other sponsors in the United States. Hence, to understand the impact of ZTP on the duration unaccompanied minors remained in ORR custody and on their family reunification prospects, we use a second dataset that includes individual-level information on all children referred to ORR after being detained by Border Patrol between October 2015 and October 2019. The sample consists of 208,000 minors between the ages of 0 and 18, with the youngest children only a few months old. The vast majority come from Guatemala (45.2%), Honduras (25.6%), and El Salvador (22.7%), followed by Mexico (2.6%) and other countries (3.9%). Over two-thirds of the study population are males, and approximately 12,200 children had not been reunited with their families by the end of the study period. Panel B in Table 1 shows how the number of unaccompanied minors referred to ORR rose by 11 percent during ZTP, their time in ORR custody lengthened from 41.6 days to 64.2 days, and the share being reunified with family dropped from 95 to 83 percent.

Additional Data

We gather data from various sources to account for time-varying country-specific traits likely altering the volume of unaccompanied minors apprehended at the border — namely, push factors. To account for the role played by violence in the origin countries, we obtain information on homicide rates from the United Nations Office on Drugs and Crime country profiles. In addition, we use Human Development Index (HDI) scores published by the United Nations Development Programme in its annual report to capture economic and human capital development likely impacting children's opportunities in each of these countries. We also gather data on UNICEF's infant and child mortality rates to address differences in access to critical health care and social services in each country, as well as information on each country's employment-to-population ratio from the World Bank to account for labor market conditions and economic opportunities at each origin. The bottom of Panel A in Table 1 presents descriptive statistics for these variables. Throughout the study period, the average homicides rate remained above 33 homicides per 100,000 population, placing the Northern Triangle and Mexico among the most violent regions in the world. In terms of economic opportunities, only 60 percent of the total working-age population were employed, and the average HDI score stood at 0.68, placing the region in the medium human development category. In consonance, infant and child mortality rates stood above the average for Latin America and the Caribbean.

Finally, to account for the role of border enforcement, which can undoubtedly impact apprehensions, we gather data on the annual budget assigned by Congress to US Border Patrol for border security operations. In addition, we collect information on the level of interior immigration enforcement and sanctuary policies in the localities where children are held while in ORR custody, which we use when modeling minors’ placement in a particular program, as well as their duration in ORR custody and family reunification odds.⁹

Methodology

Our main goal is to learn about the impact that the Trump administration's ZTP might have had on the volume of apprehended unaccompanied minors through family separations, as well as on the duration of children's custodial arrangements and their family reunification prospects. While the descriptive statistics in the prior section address some of these questions, they fail to account for other child, policy, geographic, and temporal traits likely impacting these outcomes. In this section, we address that shortcoming.

Unaccompanied Minor Apprehensions During ZTP

We start by assessing the role of the ZTP in boosting the ranks of unaccompanied minors through family separation by estimating the following benchmark model:

\begin{aligned} l o g y_{b p s, c, t} = α + β_{1} Z T P_{t} \times N T_{c} + β_{2} P o s t Z T P_{t} \times N T_{c} + β_{3} N T_{c} \\ + X_{c, t} β_{4} + β_{5} B P B u d g e t_{t} + γ_{b p s} + γ_{t} + t r e n d_{c} + ε_{b p s, c, t}, \end{aligned}

(1)

where

l o g y_{b p s, c, t}

stands for the log of unaccompanied minors apprehended in Border Patrol sector bps, while migrating from country of origin c, during month-year t. This level of aggregation allows us to account for ZTP's distinct impact on children from the Northern Triangle due to the regulation of those flows, relative to those from Mexico, and for each border sector's location and immigration enforcement history. To estimate the impact of ZTP in shaping apprehensions of unaccompanied minors, we distinguish three periods: (1) an initial period preceding the implementation of the policy between October 2013 through March 2018; (2) a second period (

Z T P_{t}

), from April 2018 through June 2018, during which the policy was in place; and (3) a final period (

P o s t Z T P_{t}

) spanning from President Trump's executive order placing family separations on hold through the end of our sample in December 2019. The period preceding the policy enactment is used as reference. We then interact the dummies indicative of each period with an indicator for whether the children originated from the non-contiguous nations in the Northern Triangle. Following the reclassification of separated children as unaccompanied and the distinct treatment of those minors based on the 2008 TVPRA, the coefficients on these interaction terms inform about changes in the volume of unaccompanied minors due to family separations taking place under ZTP.

Equation (1) also accounts for migration push factors, as captured by each origin country's homicide rates, employment-to-population ratios, HDIs, and youth mortality rates included in vector $X_{c, t}$ . In addition, we incorporate information on the monthly average of the US Border Patrol fiscal year budget ( $B P B u d g e t_{t}$ ), as well as month-year and border sector fixed effects to account for unobserved period-specific and time-invariant Border Patrol sector traits impacting apprehensions. Finally, we include country-specific time trends to control for any remaining time-varying unobservable factors in migrants’ countries of origin shaping their flows. Standard errors are clustered at the (Border Patrol sector $\times$ country of origin) level to account for variation in apprehensions of unaccompanied minors by nationality and Border Patrol sector.¹⁰

Identification hinges on the ZTP activation being uncorrelated with changes in the volume of unaccompanied minors from the Northern Triangle, relative to Mexico, other than through ongoing family separations. Reverse causality — understood as changes in migrants’ countries of origin being responsible for the adoption of ZTP — appears unlikely. Based on the Presidential Memorandum of April 6, 2018,¹¹ the tightening of border apprehensions and prosecutions under ZTP was motivated by an increasing backlog in immigration courts and the supervised release of unauthorized migrants with pending cases. In addition, the attorney general's memorandum directing the adoption of ZTP along the southwest border issued on the same day alluded to the overall rise in unauthorized crossings, without any mention of distinctions by nationality.¹²

However, we also consider that ZTP might have affected the country-of-origin composition of apprehended minors through channels other than family separations. For instance, ZTP could have deterred migrants from certain countries from attempting to cross the border based on their differential treatment under the policy. Given that TVPRA referred minors from non-contiguous countries to ORR custody, children from the Northern Triangle might have been discouraged from crossing into the United States. However, in that case, our estimates of the impact of ZTP would be downward biased, given that almost all separated families came from the Northern Triangle.

In addition to addressing some of these concerns through the inclusion of county/month fixed effects and country-specific trends that account for potentially unobserved factors driving changes in the composition of unaccompanied minors apprehended during the three-month period that the policy lasted, we conduct an event-study analysis to check on our two identifying assumptions. Specifically, we estimate the following event study model:

\begin{aligned} l o g y_{b p s, c, t} = α + \sum_{t = - 7}^{- 2} τ_{t} \cdot 1 (Z T P_{t} = 1) \times N T_{c} + \sum_{t = 0}^{7} ρ_{t} \cdot 1 (Z T P_{t} = 1) \times N T_{c} \\ + β_{1} N T_{c} + X_{c, t} β_{2} + β_{3} B P B u d g e t_{t} + γ_{b p s} + γ_{t} + t r e n d_{c} + ε_{b p s, c, t} \end{aligned}

(2)

where the indicator function

1 (Z T P_{t} = 1)

represents the

t^{t h}

month before or after the implementation of ZTP. Pre-existing differences in apprehensions between children from the Northern Triangle and Mexico are captured by the coefficients in vector

τ_{t}

. In turn, the coefficients in vector

ρ_{t}

capture the differential impact of ZTP on apprehensions of unaccompanied minors by origin up to six months after its activation.¹³ This approach enables us to gauge if, after controlling for factors included in equation (1), apprehensions of unaccompanied minors from the Northern Triangle already differed from those of Mexico prior to the activation of ZTP, in which case one could argue the policy was implemented in response to that difference or, rather, only started to diverge following the policy adoption.

Duration in Custody and Family Reunification During ZTP

We next examine how the activation of ZTP might have altered the amount of time minors were held in ORR custody and their family reunification prospects due to the administration's lack of planning for large-scale family separations. When examining how ZTP might have impacted the duration of time that children were in ORR custody, we have to adequately address right censoring in our data — namely, the fact that for some children, a discharge is never observed during our study period — as well as the nonnormal distribution of duration data. To that end, we use a semiparametric Cox proportional hazard model, which enables us to quantify the impact of ZTP on the discharge hazard of a child based on personal traits and characteristics of the locality where children were placed as follows:

\begin{aligned} λ (t | W_{i, c, t}) = λ_{0} (t) \exp (W_{i, c, t} θ) \\ = λ_{0} (t) \exp (β_{1} Z T P_{t} + β_{2} P o s t Z T P_{t} + X_{i} β_{3} + Z_{c, t} β_{4} + γ_{c} + γ_{m} + t r e n d_{c}), \end{aligned}

(3)

where

λ (t | W_{i, c, t})

is the instantaneous failure rate at time t (the “hazard”) for the i^th child placed in county c. The advantage of the Cox proportional hazard model is that it makes no assumptions about the shape of the hazard over time, which could increase, decrease, or do both at distinct periods. The main assumption is that covariates multiplicatively shift the baseline hazard, but we do not need to make assumptions about

λ_{0} (t)

. Our interest is on how ZTP impacted the discharge hazard, specifically whether it lowered it, lengthening the duration in ORR custody. To that end, vector

W_{i, c, t}

includes the temporal dummies indicative of when ZTP was in place and the post-policy period. The model also accounts for child-specific traits potentially affecting the placement decision, such as the minor's gender, age, and origin, through vector

X_{i}

. In addition, we include information on the enforcement climate in the county, as captured by its level of interior immigration enforcement and the presence of a sanctuary policy (

Z_{c, t}

). They may alter the willingness of family and friends to come forward,¹⁴ extending the duration in ORR custody and postponing family reunification. Finally, we include a time trend, as well as county and month fixed effects and, in alternative model specifications, country-specific time trends.¹⁵

We also examine whether the observed impact of ZTP on the discharge hazard pre-dated the policy adoption. To that end, we conduct the following event study:

λ (t | W_{i, c, t}) = λ_{0} (t) \exp (\sum_{t = - 7}^{- 2} τ_{t} \cdot 1 (Z T P_{t} = 1) + \sum_{t = 0}^{7} ρ_{t} \cdot 1 (Z T P_{t} = 1) + X_{i} β_{1} + Z_{c, t} β_{2} + γ_{c} + γ_{m} + t r e n d_{c}),

(4)

where the

τ

coefficients capture differences in discharge rates between treated and control groups predating the policy adoption, and the

ρ

coefficients do the same for the post-ZTP period. In both instances, period (

t - 1

) is used as reference.

Subsequently, we explore how ZTP affected children's family reunification prospects. During our study period, close to 94 percent of minors were reunified with family members. Still, more than 12,000 children aged out or were either removed, ran away, or discharged to immigration or law enforcement agencies. Because our data are right-censored, we do not observe the type of discharge some children ultimately experience. Furthermore, among those discharged, some are reunified with family, whereas others are not. To allow for the multiple failure modes, we estimate the following competing-risks model:

\begin{aligned} {\bar{h}}_{1} (t | W_{i, c, t}) = {\bar{h}}_{1, 0} (t) \exp (W_{i, c, t} θ) \\ = {\bar{h}}_{1, 0} (t) \exp (β_{1} Z T P_{t} + β_{2} P o s t Z T P_{t} + X_{i} β_{3} + Z_{c, t} β_{4} + γ_{c} + γ_{m} + t r e n d_{c}), \end{aligned}

(5)

where

{\bar{h}}_{1} (t | W_{i, c, t})

stands for the probability of family reunification for the

i

^th child in county c happening before time t in the presence of competing risks (alternative discharge types). Similar to the Cox proportional hazards model, it is assumed that the covariates multiplicatively shift the baseline sub-hazard

{\bar{h}}_{1, 0} (t)

, which is left unspecified for greater flexibility. Likewise, the vector of controls,

W_{i, c, t}

, includes the same controls as in equation (3).

Finally, we estimate an event study similar to the one specified in equation (4) using the competing-risks model to gauge if the estimated policy impacts on family reunification pre-dated the policy adoption.

Results

Unaccompanied Minor Apprehensions During ZTP

Table 2 displays the results from estimating various model specifications of equation (1). Regardless of the controls included, apprehensions of unaccompanied minors from the Northern Triangle, who are not subject to expedited removal as Mexicans are, significantly rose during the implementation of ZTP. In our baseline specification, we observe a 32 percent increase in apprehensions of unaccompanied minors from the Northern Triangle, compared to those from Mexico, following the activation of ZTP. This estimate rises to 48 percent when we account for migration push factors and Border Patrol's budget in column (3). Furthermore, the differential impact of ZTP on apprehensions of unaccompanied minors from the Northern Triangle, compared to those of Mexicans, seems to dissipate after the policy's suspension.

Table 2.

ZTP and Unaccompanied Minors Apprehended by US Border Patrol, 2013–2019.

Model specification:	(1)	(2)	(3)
Northern Triangle (NT)	−1.145***	−0.771	−0.771
	(0.357)	(0.806)	(0.806)
ZTP period $\times$ NT	0.316**	0.478***	0.478***
	(0.127)	(0.141)	(0.141)
Post-ZTP period $\times$ NT	0.197	0.277	0.277
	(0.175)	(0.182)	(0.182)
Country of origin time-varying controls	No	Yes	Yes
Border Patrol budget for the fiscal year	No	No	Yes
Month-year & Border sector fixed effects	Yes	Yes	Yes
Country-specific time trends	Yes	Yes	Yes
Observations	2,700	2,700	2,700
R-squared	0.645	0.672	0.672

Note: The period is October 2013 through December 2019. Data on unaccompanied minor apprehensions are aggregated at the (Border Patrol sector $\times$ country of origin $\times$ month-year) level. All continuous variables are in logs. All regressions include a constant, and month-year and border sector fixed effects. The pre-ZTP period is used as reference to the ZTP and post-ZTP periods. Country of origin's time-varying controls include the homicide rate, employment-to-population ratio, HDI, and mortality rates for children 0 to 5 and 5 to 14 years old. Standard errors in parentheses are clustered at the (Border Patrol sector $\times$ country of origin) level. *p < .1; **p < .05; ***p < .01.

As noted earlier, an important limitation of the empirical strategy used in Table 2 refers to the possibility of pre-ZTP differential trends in apprehensions of unaccompanied minors from the Northern Triangle relative to those from Mexico. We estimate the event study model in equation (2) to gauge if that was the case.

Figure 4 displays the estimated coefficients ( $τ_{t}$ and $ρ_{t}$ ), along with 95 percent confidence intervals. All estimates for the periods preceding the activation of the ZTP are close to zero, strongly supporting the lack of differential pre-trends in apprehensions of unaccompanied minors from the Northern Triangle versus Mexico. In addition, we observe a clear break in this trend immediately after the activation of ZTP — a difference that remains statistically different from zero for the duration of the policy. Altogether, the results in Figure 4 support the assumption of parallel trends in apprehensions of unaccompanied minors from the Northern Triangle and Mexico prior to ZTP and, in turn, the interpretation of the policy impacts in Table 2 as causal.

Figure 4.

Event study for apprehensions of unaccompanied minors relative to ZTP activation.

Duration in Custody and Family Reunification During the ZTP

After assessing the impact of ZTP on the volume of unaccompanied minors, we turn to examine how the policy implementation might have altered the duration in ORR custody and the children's prospects of being reunified with family. Using a Cox proportional hazard model, we first estimate the discharge hazard ratios during and after ZTP relative to the pre-policy period. Results from estimating the model in equation (3) are shown in Table 3. We estimate various model specifications that progressively include more controls, and our findings prove robust to their inclusion. Based on the most complete model specification in column (4), the discharge rate of unaccompanied children dropped by 38 percent during ZTP, but recovered in the post-ZTP period.

Table 3.

Cox Proportional Hazard Model for Discharge From ORR Custody, 2015–2019.

	Discharge hazard (Robust SE)	Discharge hazard (Robust SE)	Discharge hazard (Robust SE)	Discharge hazard (Robust SE)
Model specification	(1)	(2)	(3)	(4)
ZTP period	0.508***	0.509***	0.616***	0.624***
	(0.005)	(0.005)	(0.007)	(0.007)
Post-ZTP period	0.810***	0.812***	1.071***	1.049***
	(0.004)	(0.004)	(0.011)	(0.011)
Child demographic controls	Yes	Yes	Yes	Yes
Month fixed effects	Yes	Yes	Yes	Yes
County time-varying controls	No	Yes	Yes	Yes
Time trend	No	No	Yes	Yes
Country-specific time trends	No	No	No	Yes
Observations	200,840	200,840	200,840	200,840

Note: The period is October 2015 through October 2019. The table presents the discharge hazard ratios from a Cox proportional hazards model that examines time (measured in days) to discharge from ORR custody. The pre-ZTP period is used as reference to the ZTP and post-ZTP periods. Child demographic controls include gender, age, and country of origin. County time-varying controls include information on the county's level of interior immigration enforcement and the presence of a sanctuary policy. Robust standard errors in parentheses. *p < .1; **p < .05; ***p < .01.

To better understand the long-term impact of the policy on children's duration in ORR custody, Figure 5 plots the corresponding Kaplan-Meier survival rates. It shows that two months after being transferred to ORR custody, about 38 percent of unaccompanied minors were still under the agency's watch, relative to 20 percent before ZTP. Even after the policy's suspension, approximately 25 percent of the children remained in ORR custody two months after their transfer. In fact, the probability of remaining in ORR custody before, during, and after ZTP only converged to zero eight months after the children were initially transferred to ORR.

Figure 5.

Length of time in ORR custody Kaplan-Meier survival rates relative to ZTP activation.

We also verify that pre-existing trends are not driving our results by estimating the event study in equation (4). Figure 6 presents the 95 percent confidence intervals for ORR discharge hazards for the six months before and after the activation of ZTP. The results show that none of the coefficients in the pre-treatment period are statistically different from zero. Further, we observe a clear break starting in period t, with the discharge rate of unaccompanied children dropping by approximately 40 percent. The impact only starts to dissipate once ZTP is terminated but is not immediately reversed. These results support the causality of our estimates in Table 3.

Figure 6.

Event study for ORR discharge hazards relative to ZTP activation.

We next explore how ZTP impacted family reunification. Table 4 displays the family reunification hazards from various model specifications of the competing risks model described in equation (5). As with prior outcomes, the policy impact appears robust to the inclusion of additional controls. The estimates from the most complete model specification in column (4) reveal that the odds of family reunification during ZTP were 49 percent lower than prior to the policy's activation, even though they improved after the policy's suspension.

Table 4.

Competing-Risks Model for Family Reunification, 2015–2019.

	Sub-hazard (Robust SE)	Sub-hazard (Robust SE)	Sub-hazard (Robust SE)	Sub-hazard (Robust SE)
Model specification	(1)	(2)	(3)	(4)
ZTP period	0.418***	0.418***	0.503***	0.512***
	(0.004)	(0.004)	(0.006)	(0.006)
Post-ZTP period	0.818***	0.818***	1.073***	1.055***
	(0.004)	(0.004)	(0.011)	(0.011)
Child traits	Yes	Yes	Yes	Yes
Month fixed effects	Yes	Yes	Yes	Yes
County level traits	No	Yes	Yes	Yes
Time trend	No	No	Yes	Yes
Country of origin time trends	No	No	No	Yes
Observations	200,840	200,840	200,840	200,840

Note: The period is October 2015 through October 2019. The table presents the sub-hazard ratios from a competing-risks model that analyzes time to reunification after a minor is discharged from ORR custody while considering alternative forms of discharge as competing events. The pre-ZTP period is used as reference to the ZTP and post-ZTP periods. Child traits include gender, age, and country of origin. County level traits include information on the county's level of interior immigration enforcement and the presence of a sanctuary policy. Robust standard errors in parentheses. *p < .1; **p < .05; ***p < .01.

Figure 7 underscores the catastrophic consequences of ZTP on these children's family reunification odds by displaying cumulative reunification incidence curves. It shows that only 50 percent of the children had been reunified with their families two months after being placed in ORR custody, compared to close to 80 percent before ZTP. Perhaps most tragically, one year after being placed under ORR's watch, family reunification remained almost 20 percent less likely than before ZTP was implemented.

Figure 7.

Cumulative reunification incidence by period relative to ZTP activation.

As a final note, we consider the possibility that pre-treatment trends could be driving our findings regarding the reunification of unaccompanied minors with their families. Using the competing-risks model, we conduct an event study similar to equation (4) and find that while the pre-treatment estimates are close to zero, there is a sharp decline while ZTP is active (Figure 8). Once more, these results support our estimates as causal.

Figure 8.

Event study for family reunification sub-hazards relative to ZTP activation.

Heterogeneous Impacts by Age

Thus far, we have shown how the implementation of ZTP had deleterious impacts on the discharge and family reunification rates of unaccompanied minors. Of particular concern for many was the disproportionate damage that separations might have on very young children, whose ability to process and digest what is happening to them might be particularly limited. For that reason, we next explore heterogeneity in the policy impacts on tender age children (12 years old and younger) and their teenage counterparts (13–17 years old). Tables 5 and 6 show the results from that exercise.

Table 5.

Heterogeneous Effects — ORR Discharge Hazards by Age Group, 2015–2019.

	Discharge hazard (Robust SE)
	(1)	(2)
Tender age (0–12)
ZTP period	0.537***	0.540***
	(0.010)	(0.010)
Post-ZTP period	0.868***	0.839***
	(0.014)	(0.014)
Non-tender age (13–17)
ZTP period	0.627***	0.636***
	(0.007)	(0.010)
Post-ZTP period	1.106***	1.087***
	(0.011)	(0.011)
Child traits	Yes	Yes
Month fixed effects	Yes	Yes
County level traits	No	Yes
Time trend	No	No
Country of origin time trends	No	No
Observations	200,840	200,840

Note: The table presents the discharge hazard ratios from a Cox proportional hazards model that examines time (measured in days) to discharge from ORR custody by children's age group. The pre-ZTP period is used as reference to the ZTP and post-ZTP periods. Child demographic controls include gender, age, and country of origin. County time-varying controls include information on the county's level of interior immigration enforcement and the presence of a sanctuary policy. Robust standard errors in parentheses. *p < .1; **p < .05; ***p < .01.

Table 6.

Heterogeneous Effects — Family Reunification Sub-Hazards by Age Group, 2015–2019.

	Reunification sub-hazard(Robust SE)
	(1)	(2)
Tender age (0–12)
ZTP period	0.263***	0.267***
	(0.007)	(0.007)
Post-ZTP period	0.845***	0.822***
	(0.014)	(0.013)
Non-tender age (13–17)
ZTP period	0.583***	0.595***
	(0.007)	(0.007)
Post-ZTP period	1.113***	1.097***
	(0.011)	(0.011)
Child traits	Yes	Yes
Month fixed effects	Yes	Yes
County level traits	No	Yes
Time trend	No	No
Country of origin time trends	No	No
Observations	200,840	200,840

Note: The table presents the sub-hazard ratios by age group from a competing-risks model that analyzes time to reunification after a minor is discharged from ORR custody while considering alternative forms of discharge as competing events. The pre-ZTP period is used as reference to the ZTP and post-ZTP periods. Child traits include gender, age, and country of origin. County level traits include information on the county's level of interior immigration enforcement and the presence of a sanctuary policy. Robust standard errors in parentheses. *p < .1; **p < .05; ***p < .01.

A few results are worth discussing. First, although ZTP substantially lowered discharge rates among both groups of children, its impact was larger among tender age kids, for whom their discharge rate dropped by 10 percentage points more than for their teenage counterparts, that is, by 46 percent versus 36 percent, respectively. Secondly, the impacts were long lasting for tender age children, whose discharge rates remained lower during the post-ZTP period under study, while it recovered for teenage UACs.

We next explore the heterogeneous impacts of the policy on family reunification rates of tender-age and teenage children in Table 6. Again, a couple of results are worth highlighting. First, the family reunification rate of tender-age children during ZTP dipped to 27 percent of its rate prior to the policy's activation — a drop substantially greater than the one experienced by teenage youth. For the latter, the family reunification rate dropped to 60 percent of its pre-policy level. Secondly, even though the likelihood of family reunification for tender-age children somewhat recovered when the policy ended, reaching close to 80 percent, it remained lower than prior to ZTP. As a result, it is not surprising to find that the cumulative reunification incidence for tender-age children affected by ZTP was still more than 30 percentage points lower one year after placement in ORR custody than before the policy's adoption, as depicted in Figure 9.

Figure 9.

Cumulative reunification incidence by period and age group.

Summary and Conclusions

Using two separate datasets on apprehensions of unaccompanied minors and unaccompanied minors referred to ORR gathered through FOIA requests, we examine the impact of the Trump administration's ZTP on boosting the ranks of unaccompanied minors through family separation, as well as on children's duration in ORR custody and their family reunification prospects.

The findings reveal the very damaging impact of the policy in all respects. First, the adoption of ZTP resulted in a 48 percent increase in apprehensions of unaccompanied minors due to family separations, particularly among Central American families. Event studies further confirm how the estimated differential policy impact did not predate its adoption and, most worrisomely, its effect lasted long after its suspension by President Trump in June 2018. Secondly, the program also altered the duration children remained in ORR custody by lowering their discharge rate from the agency's watch by 38 percent. As a result, two months after being transferred to ORR custody, almost double the share of children remained under the agency's watch than before the policy's adoption (38% vs. 20%). Lastly, the odds of family reunification during ZTP dropped by 49 percent relative to the period prior to the policy's activation. Consequently, two months after being placed in ORR custody, only 50 percent of the children were successfully reunified with family, as opposed to 80 percent prior to ZTP. In fact, under ZTP, the odds of being reunified with family were 20 percentage points lower one year after the children were placed in ORR custody than before the implementation of the policy. These impacts were substantially stronger among tender-age children, highlighting the harmful effects of the policy.

In sum, ZTP appears to have had very damaging impacts on children and the treatment they received. Given the potential socio-emotional, behavioral, and health implications of separation, as captured by increased anxiety, depression, post-traumatic stress, and attention-deficit disorders that clinicians have observed in detained children,¹⁶ further attention to the implications of adopting harsh immigration enforcement policies is well-warranted.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Catalina Amuedo-Dorantes

José R. Bucheli

Notes

Appendix

Table A1.

Relevant Legislation.

Legislation	Description
Flores Settlement Agreement (1997)	Lawsuit filed in 1985 by migrant children challenging government procedures related to their detention, treatment, and release. According to the agreement, children under government custody have to be held in “least restrictive setting.”
Homeland Security Act (2002)	In compliance with Flores Settlement Agreement, the law transferred authority over the processing and care of unaccompanied minors from the extinct Immigration and Naturalization Service (INS) to the Office of Refugee Resettlement (ORR) under the Department of Health and Human Services.
Trafficking Victims Protection Reauthorization Act (2008)	Unaccompanied migrant children from contiguous countries are repatriated if no grounds for asylum. Unaccompanied children from non-contiguous countries are transferred over to ORR custody within 72h of apprehension. ORR places children in least restrictive setting.
Zero-Tolerance Policy (2018)	All adult unauthorized border crossers criminally prosecuted regardless of intention to apply for asylum or whether traveling with minors. Adults are detained in federal criminal facilities, children are separated from their families, reclassified as ‘unaccompanied minors,’ and referred to ORR custody.

References

American Civil Liberties Union. 2018. “Family Separation by the Numbers.” https://www.aclu.org/issues/family-separation.

Amuedo-Dorantes

Bucheli

J. R.

. 2023. “Immigration Policy and Hispanic Representation in National Elections.” Journal of Policy Analysis and Management. https://doi.org/10.1002/pam.22444.

Amuedo-Dorantes

Puttitanun

. 2016. “DACA and the Surge in Unaccompanied Minors at the US-Mexico Border.” International Migration 54 (4): 102–17. https://doi.org/10.1111/imig.12250.

Capps

Meissner

Soto

A. G. R.

Bolter

Pierce

. 2019. “From Control to Crisis: Changing Trends and Policies Reshaping U.S.-Mexico Border Enforcement.” Migration Policy Institute .

Carey

2018. “A Troubling Prognosis for Migrant Children in Detention: ‘The Earlier They’Re Out, the Better’.” The New York Times , June 18, 2018. https://www.nytimes.com/2018/06/18/health/migrant-children-mental-health.html.

Clemens

M. A.

2021. “Violence, Development, and Migration Waves: Evidence from Central American Child Migrant Apprehensions.” Journal of Urban Economics 124: 103355. https://doi.org/10.1016/j.jue.2021.103355.

Corradini

Kringen

J. A.

Simich

Berberich

Emigh

. 2018. “Operation Streamline: No Evidence That Criminal Prosecution Deters Migration.” Vera Institute of Justice . https://www.vera.org/downloads/publications/operation_streamline-report.pdf.

Flynn

2018. “Federal Judge Denies Trump Administration’s Request to Indefinitely Detain Families.” The Washington Post , July 18, 2018. https://www.washingtonpost.com/news/morning-mix/wp/2018/07/10/federal-judge-denies-trump-administrations-request-to-indefinitely-detain-families/.

Human Rights Watch. 2018. “Q&A: Trump Administration’s “Zero-Tolerance” Immigration Policy.” https://www.hrw.org/news/2018/08/16/qa-trump-administrations-zero-tolerance-immigration-policy.

10.

Interagency Task Force on the Reunification of Families. 2021. “Initial Progress Report.” U.S. Department of Homeland Security. https://www.dhs.gov/publication/family-reunification-task-force-progress-reports.

11.

Jones

Obser

Podkul

. 2017. “Betraying Family Values: How Immigration Policy at the United States Border Is Separating Families.” Lutheran Immigration and Refugee Service (LIRS), the Women’s Refugee Commission (WRC), and Kids in Need of Defense (KIND). https://supportkind.org/wp-content/uploads/2017/03/BetrayingFamilyValues_Feb2017.pdf.

12.

Linton

J. M.

Griffin

Shapiro

A. J.

. 2017. “Detention of Immigrant Children.” Pediatrics 139 (5), https://doi.org/10.1542/peds.2017-0483.

13.

Ryo

Humphrey

. 2021. “Children in Custody: A Study of Detained Migrant Children in the United States.” UCLA Law Review 136: 136–211. https://www.uclalawreview.org/children-in-custody-a-study-of-detained-migrant-children-in-the-united-states/.

14.

UNICEF-UNHCR. 2020. “Families on the Run.” https://familiesontherun.org.

15.

U.S. Congressional Research Service. 2021. “The Trump Administration’s “Zero Tolerance” Immigration Enforcement Policy.” Vol. R45266. https://crsreports.congress.gov/product/pdf/R/R45266.

16.

U.S. Customs and Border Protection. 2021a. “Southwest Land Border Encounters (By Component): Fiscal Year 2021 Stats.” https://www.cbp.gov/newsroom/stats/southwest-land-border-encounters-by-component.

17.

U.S. Customs and Border Protection. 2021b. “U.S. Border Patrol Total Monthly UC Encounters by Sector (FY 2010 - FY 2020).” Vol. 1517–0721. https://www.cbp.gov/document/stats/us-border-patrol-total-monthly-uc-encounters-sector-fy-2010-fy-2020.

18.

U.S. Department of Health and Human Services. 2020. “Communication and Management Challenges Impeded HHS’s Response to Zero-Tolerance Policy.” Vol. OEI-BL-18-. Office of Inspector General. https://oig.hhs.gov/oei/reports/oei-BL-18-00510.pdf.

19.

U.S. Department of Justice. 2021. “Review of the Department of Justice’s Planning and Implementation of Its Zero Tolerance Policy and Its Coordination with the Departments of Homeland Security and Health and Human Services.” Vol. 21–028. Office of Inspector General. https://oig.justice.gov/sites/default/files/reports/21-028_0.pdf.