From Volume 92, Number 1 (November 2018)
A National Study of Immigration Detention in the United StateS[*]
Emily Ryo[†] and Ian Peacock[‡]
Amidst growing reports of abuses and rights violations in immigration detention, the Trump administration has sought to expand the use of immigration detention to facilitate its deportation policy. This study offers the first comprehensive empirical analysis of U.S. immigration detention at the national level. Drawing on administrative records and geocoded data pertaining to all noncitizens who were detained by U.S. Immigration and Customs Enforcement in fiscal year 2015, we examine who the detainees are, where they were held, and what happened to them.
The bulk of the detained population consisted of men (79%) and individuals from Mexico, El Salvador, Guatemala, and Honduras (together, 89%). Over 59,000, or about 17%, of the detainees were juveniles under the age of eighteen. Every state in the United States had one or more facility, with Texas and California having the highest number of facilities and detainees. Detention in privately operated facilities and in remote locations was common. We analyze three key detention outcomes: detention length, inter-facility transfers, and facility-related grievances. The average detention length for adults released in fiscal year 2015 was thirty-eight days, though tens of thousands were detained for many months or years. A majority of these detainees experienced one or more inter-facility transfers, many involving movements across cities, states, and federal judicial circuits. In fiscal year 2015, the Detention Reporting and Information Line received over 48,800 facility-related grievances, a majority of which concerned issues pertaining to access to legal counsel and basic immigration case information.
We find that detention outcomes vary significantly across facility operator types (private versus non–private) and facility locations (within or outside of major urban areas). Specifically, our multivariate regression analyses show that confinement in privately operated facilities is associated with significantly longer detention and a higher number of grievances. We find a similar pattern of results for confinement in facilities located outside of major urban areas. On the other hand, confinement in privately operated facilities, and confinement in facilities located outside of major urban areas, respectively, are associated with lower risks of inter-facility transfers. These findings provide an important foundation for ongoing public discourse and policy discussions on the expanded use of detention as an immigration enforcement strategy.
In Zadvydas v. Davis, a case about the constitutionality of indefinite immigration detention, the U.S. Supreme Court declared: “Freedom from imprisonment—from government custody, detention, or other forms of physical restraint—lies at the heart of the liberty that [the Due Process] Clause protects.” As courts have noted, imprisonment implicates core due process issues because imprisonment not only engenders the loss of freedom of physical movement, but also inflicts deep social stigma and other enduring adverse consequences on the individual.
Every day, tens of thousands of noncitizens in immigration proceedings face this deprivation of personal liberty through immigration detention. The U.S. Department of Homeland Security (“DHS”), rather than the U.S. Department of Justice (“DOJ”), currently operates the largest confinement and supervised release program in the United States. Under the Trump administration, the number of people booked into the custody of U.S. Immigration and Customs Enforcement (“ICE”) through its interior enforcement program has increased steeply. This is not surprising given the administration’s announcement to employ detention as the default immigration enforcement strategy. In accordance with this strategy, the Trump administration plans to build additional detention facilities across the country to accommodate the expected rise in the detainee population. The Trump administration has also sought to lower the detention standards in order to make the facility contracts more palatable to third-party contractors.
This shift in policy is taking place amidst growing reports that have documented a host of due process violations and human rights abuses in immigration detention. For example, these reports have drawn public attention to the rising number of deaths and suicides in detention, dangerous and substandard medical care, sexual and physical abuse, exploitative labor practices, and lack of adequate access to legal counsel, among many other issues. Due to data scarcity, however, many of these investigative reports have been relatively limited in their scope and primarily focused on specific issues, regions of the country, or subpopulations. The same is largely true of scholarly research on immigration detention in the United States. Fundamental questions thus remain at the national level about the detained population, the facilities where the detainees are confined, and their experiences and outcomes.
This study offers, for the first time, a comprehensive empirical analysis of U.S. immigration detention at the national level. Drawing on administrative records pertaining to all individuals who were in ICE custody in fiscal year 2015, as well as geocoded data and records of grievances relating to the facilities in which the detainees were confined, we examine who the detainees are, where they were held, and what happened to them. We also examine factors that predict variations in the following key detention outcomes: detention length, the number of inter-facility transfers that the detainees experienced during detention, and the number of grievances filed against detention facilities. Detention length and grievances capture basic aspects of detention experiences and outcomes. Inter-facility transfers constitute another important—albeit largely overlooked—measure of what happens to individuals in detention. Transfers warrant a special scrutiny because they can substantially hinder access to legal representation, sever family ties and community support, and separate detainees from the evidence needed in their court proceedings.
In analyzing these key detention outcomes, we pay special attention to two features of the U.S. detention system that have become a focal point of growing concern among advocates, scholars, and policymakers. The first is the expanding role of private companies in the U.S. detention system. According to a recent government report, 65% of the average daily detainee population as of September 2016 were confined in facilities operated by private, for-profit contractors. The second prominent feature of the current U.S. detention system relates to the location—or more precisely, the relative remoteness—of many of the detention facilities. In short, we assess whether the detention outcomes of interest in this study are related to confinement in privately operated facilities and in facilities that are located outside of major urban areas.
The remainder of this Article proceeds in three major parts. Part I provides the basic legal, political, and research context for understanding immigration detention as it has evolved over time and as it stands now. Part II describes the data we analyze in this Article. We obtained and merged three major datasets to conduct our analyses. The primary dataset comes from records that ICE provided to the Transactional Records Access Clearinghouse (“TRAC”) pursuant to the Freedom of Information Act (“FOIA”). This dataset consists of longitudinal information on each individual detained by ICE during fiscal year 2015. The second dataset is a compilation of geocoded records that allow us to examine distances to and from detention facilities and other locations of interest in this study. The third dataset consists of records that Human Rights Watch obtained through FOIA from ICE on the complaints and grievances that detainees and other stakeholders submitted involving the detention facilities. Part III presents our key empirical findings.
Taken together, our findings constitute an important first step toward understanding the possible structural determinants of detention experiences and outcomes. More generally, our findings provide a critical empirical foundation for future research and policy debates on the expanded use of detention as an immigration enforcement strategy.
We begin with a brief discussion of the current political context of immigration detention. The modern era of immigration detention in the United States can be traced to the enactment of two laws in 1996: The Antiterrorism and Effective Death Penalty Act (“AEDPA”) and the Illegal Immigration Reform and Immigrant Responsibility Act (“IIRIRA”). The AEDPA broadened the list of crimes defined as an aggravated felony and expanded the types of offenses (beyond aggravated felonies) that trigger mandatory detention.
A few months later, Congress enacted the IIRIRA to further expand the use of immigration detention, including by broadening the categories of noncitizens subject to mandatory detention. Aggravated felonies now include offenses that are neither aggravated nor a felony under criminal statutes, such as a simple battery or shoplifting conviction. Further, the category of individuals subject to mandatory detention now extends not only to noncitizens with criminal convictions, but also to certain classes of arriving aliens.
As expected, the number of noncitizens detained by immigration authorities during the post–1996 period rose steadily and dramatically. In 1994, an average of 6,785 noncitizens were detained on any given day. By 2014, that daily average had surpassed 33,200 (see Figure 1). The average length of detention also increased during this time period, from twenty-six days to thirty days. The DHS’s budget for fiscal year 2017 estimated an average rate of $126.46 per day for adult detention beds and an average rate of $161.36 per day for family detention beds.
The Trump administration’s immigration enforcement policy seeks to bring dramatic changes in the use and operation of immigration detention. First, signaling a sharp departure from the Obama administration’s policy of prioritizing noncitizens with criminal convictions for removal, the Trump administration’s plan targets virtually all unauthorized immigrants regardless of whether they have been convicted of a crime. This policy shift has already resulted in increased removal rates of noncitizens without criminal convictions. Second, the Trump administration seeks to detain all noncitizens apprehended at the border pending their immigration proceedings.
To implement these plans, President Trump has called for an allocation of “all legally available resources” for the immediate construction, operation, and control of detention facilities near the border with Mexico, or for the establishment of contracts for such facilities. In June 2017, Thomas Homan, the ICE Acting Director, stated that his budget for the fiscal year 2018 included nearly $4.9 billion to expand the average daily detained population to over 51,000. According to the latest government statistics, the overall book-ins to ICE detention declined in fiscal year 2017 due to a decline in book-ins resulting from border apprehensions. However, the initial book-ins resulting from ICE’s interior enforcement programs were 42% higher between January 20, 2017 and September 20, 2017 (the period after President Trump took office), compared to the same time period in fiscal year 2016.
B.Legal and Policy Framework
There are many aspects of immigration detention that make detention indistinguishable from criminal incarceration. Yet the law considers immigration detention to be strictly civil—that is, “nonpunitive and merely preventative” in nature. Consequently, the basic legal protections that are afforded to criminal defendants, such as the right to government-appointed counsel, the privilege against self-incrimination, the ban on cruel and unusual punishment, and the right to a speedy trial, are deemed to be inapplicable in the immigration law context.
Beyond the basic principle that detention is civil, the legal framework governing immigration detention is exceedingly complex and has changed over time. Below, we offer a broad overview of the current legal and policy framework that governs basic aspects of immigration detention in the United States. More specifically, our overview focuses on issues relating to which classes of noncitizens may be detained, the locations where they may be held, and the conditions of their confinement.
A number of provisions of the Immigration and Nationality Act (“INA”) grant immigration officials the power to detain noncitizens for the purposes of immigration enforcement. First, under INA section236(a), “an alien may be arrested and detained pending a decision on whether the alien is to be removed from the United States.” Because of the permissive language of “may” used in this provision, section236(a) is often referred to as the discretionary detention provision of the INA. Noncitizens detained under this provision may be released on conditional parole (also commonly known as “release on recognizance”) or on a bond of at least $1,500. Noncitizens released under this provision, however, may be rearrested at any time at the discretion of an authorized immigration official.
Second, under what are often referred to as the mandatory detention provisions of the INA, the Attorney General must detain, with limited exceptions, certain classes of noncitizens pending their removal from the United States. These noncitizens include: (1) most “arriving aliens;” (2) noncitizens with certain criminal convictions; (3) suspected terrorists; and (4) noncitizens with final orders of removal. An “arriving alien” generally refers to an “applicant for admission,” such as an individual apprehended at the border or a port of entry. As there are no statutory limits on how long a noncitizen may be detained, and many noncitizens are detained for extended periods of time, prolonged detention has been the focus of ongoing litigation in federal courts.
As to where the detainees may be held, INA section241(g) states: “TheAttorney Generalshall arrange for appropriate places of detention for aliens detained pending removal or a decision on removal.” This provision further states:
When United States Government facilities are unavailable or facilities adapted or suitably located for detention are unavailable for rental, theAttorney Generalmay expend...amounts necessary to acquire land and to acquire, build, remodel, repair, and operate facilities (including living quarters for immigration officers if not otherwise available) necessary for detention.
One important issue related to where the detainees may be held is whether and when detainees may be transferred from one facility to another. On this question, the federal courts generally have interpreted INA section241(g) as providing the Attorney General broad discretion to transfer the detainees as he or she “deems appropriate.” Thus, noncitizens have been held where they initially were apprehended or transferred to remote and distant facilities, including ones that are outside the jurisdiction of the presiding court. In 2009, a report by the Office of Inspector General concluded that such transfers have led to “errors, delays, and confusion for detainees, their families, and legal representatives.” Around the same time, Human Rights Watch published a report raising serious concerns about the frequent transfers of large numbers of detainees to facilities that were far away from where they lived. In response, ICE informed Human Rights Watch that it intended to minimize transfers. In 2012, ICE released a policy directive restricting detainee transfers to only those deemed to be “necessary” under a specific set of circumstances. In practice, legal advocates continue to report that “often ICE does not follow its policy memorandum and instead bases the location of detention on bed space availability.”
As to the type of detention facilities used by ICE, the DHS recognized in 2009 that most immigrant detainees were confined in facilities that were “either jails operated by county authorities or detention centers operated by private contractors.” The DHS also recognized that these facilities were “largely designed for penal, not civil, detention.” The DHS announced an overhaul of this system with the creation of the Office of Detention Policy and Planning (“ODPP”). The announcement stated that ICE would move away from its reliance on penal institutions to confine immigrant detainees and instead, design “a new civil detention system.”
In 2016, in the aftermath of the DOJ’s decision to phase out private for–profit prisons from the federal correctional system, the Homeland Security Advisory Council considered whether the same phase-out ought to be implemented in the immigration detention system. The Advisory Council’s report, however, concluded: “Fiscal considerations, combined with the need for realistic capacity to handle sudden increases in detention, indicate that DHS’s use of private for-profit detention will continue.” This conclusion was the subject of a contentious debate among the Council members, resulting in almost three-fourths of the Council members concurring with one Council member’s view that the federal government should take a “measured but deliberate shift away from the private prison model.”
The treatment of detainees and their conditions of confinement are governed by a number of different ICE detention standards. Three versions of the standards are currently in use across various facilities throughout the United States, depending on the type of facility and the terms of the facilities’ respective contracts with ICE. These standards include the 2000 National Detention Standards (“NDS”), and the 2008 and 2011 Performance-Based National Detention Standards (“PBNDS”). None of these standards are legally enforceable regulations. As the U.S. Commission on Civil Rights has concluded, the nonbinding nature of these standards often means that the “facilities are not held accountable when they fail to maintain or meet these standards—at times with tragic results.”
A detailed review of the specific issues related to the conditions of confinement and the treatment of detainees is beyond the scope of this Article. As we noted earlier, however, these issues are manifold, ongoing, and broad ranging. The Office of the Inspector General underscored this point in reaching the following conclusion in its December 2017 report based on its unannounced inspections of several detention facilities: “Overall, the problems we identified undermine the protection of detainees’ rights, their humane treatment, and the provision of a safe and healthy environment.”
C.Research on Detention
The existing research on immigration detention offers another important context for our empirical analysis. The extant body of research is varied in approach and wide-ranging in focus. A longstanding body of legal scholarship has analyzed statutory, constitutional, doctrinal, and policy issues related to immigration detention, particularly mandatory detention. Complementing this legal scholarship is a relatively small, but growing, body of scholarship from multiple disciplines that illuminates the history, politics, and social realities of immigration detention. Finally, there is a nascent body of socio-legal scholarship that brings empirical analysis to bear upon legal and policy issues related to immigration detention.
A number of themes have emerged as focal points of inquiry across these related albeit disparate bodies of research. Here, we focus on the research on the political economy of detention. By political economy, we mean the confluence of political and economic forces that determine the emergence, location, and the locus of governance and power over the immigration detention infrastructure. An important point of inquiry in this area of research has been the increasing privatization of immigration enforcement, which involves the delegation of enforcement powers—a traditional government function—to private actors.
That detention has become a core topic of investigation for researchers concerned with the privatization of immigration enforcement is unsurprising, given the expanded role of for-profit companies in the construction and operation of immigration detention facilities across the United States. Due largely to data limitations, evidence is mixed on the precise role of privatized detention in shaping detention experiences and outcomes. On the one hand, Jennifer Chacón, in her recent analysis of the critiques of privatized detention, has concluded that currently, systematic evidence is lacking on whether detainees are worse off in privately operated facilities compared to public facilities. On the other hand, Denise Gilman and Luis Romero have argued that privatization fuels profit-seeking dynamics that distort decisions regarding whether to detain, where to detain, and for how long.
Research on the political economy of detention has also focused on the centrality of geopolitics in structuring detention experiences and outcomes. Alison Mountz, for example, has argued that the practice of holding detainees in remote locations that are outside and distant from urban areas leads to separation from family, community, and legal support networks that are critical to the detainees’ chances of achieving favorable case outcomes. Furthermore, Lauren Martin has argued that ICE “is acutely aware of detainees’ relationships to surrounding communities,” and considers proximity to such communities and their supporting institutions as a negative trait in selecting detention sites.
Taken together, the foregoing discussion of the key themes in detention law, policies, and research raises fundamental questions about who the detainees are, where they are held, and what happens to them once they are detained. We now turn to our analysis in which we investigate these issues using a national dataset on individuals held in immigration detention by ICE during fiscal year 2015.
II.The Current Study
We compiled a number of datasets for this study. Together, these datasets represent the most comprehensive national-level data to date on U.S. immigration detention. We discuss each of these data sources and their limitations, followed by a brief discussion of our analytical strategy.
The primary dataset consists of individual-level longitudinal data on each individual, including juveniles, detained by ICE during fiscal year 2015 (“Detention Data”). TRAC obtained the Detention Data from ICE with public records requests. To our knowledge, 2015 is the latest and the only fiscal year that the federal government has released individual-level data of this kind on immigration detention. TRAC is currently awaiting ICE’s response to its request for more recent detention data.
For each detainee, ICE generated a new record each time the detainee was booked into a facility. We treat each record as a new “stint” in detention. If an individual had been booked into a facility only once during the course of his detention, that detainee would have only one record pertaining to that detention stint. On the other hand, if an individual had been booked into a facility, subsequently transferred to another facility, and then released from the second facility on parole, that detainee would have a total of two records pertaining to two detention stints. The first record would have the release reason (“Release Type” variable in the Detention Data) of “transferred,” and the second record would have the release reason of “paroled.”
To be included in the Detention Data, the individual must have been detained at some point during fiscal year 2015, but his or her detention need not have begun nor ended in fiscal year 2015. For individuals who entered detention before fiscal year 2015, some of their records in the Detention Data pre-date fiscal year 2015. For individuals whose detention continued beyond fiscal year 2015, we do not observe what happened to them and their records are right censored. A total of 38,796 records are right censored in the Detention Data.
We undertook a number of steps to clean and prepare the data for analysis, including deleting duplicate or redundant records. The Methods Appendix describes each of these steps. Our data cleaning and preparation resulted in 734,709 records pertaining to 355,729 detainees, including juveniles and adults, and 679 detention facilities. For each detention facility in the Detention Data, we collected information from a number of sources to determine the facility’s address and whether the facility was operated by a private for-profit company. The Methods Appendix contains detailed information on our coding process.
We generated a geocoded dataset specifically for the purposes of this study (“Geocoded Data”). To create this dataset, we compiled and merged three distinct record sets. The first set of records consists of a comprehensive list of facility dyads that we produced, with each dyad consisting of pairings of all facilities involved in inter-facility transfers.
The second set of records pertains to the metropolitan statistical areas (“MSAs”) and their principal cities. The U.S. Office of Management and Budget (“OMB”) defines MSAs as consisting of “at least one urbanized area that has a population of at least 50,000.” About 85% of the U.S. population live in MSAs. Given the expansive coverage of MSAs, we sought a narrower definition of major urban areas for the purposes of this study. The largest city in each MSA is called a principal city. According to the OMB, principal cities constitute “the more significant places in each [MSA]...in terms of population and employment.” Thus, we define major urban areas in this study as the principal cities in MSAs.
The third set of records pertains to immigration attorneys who are members of the American Immigration Lawyers Association (“AILA”). AILA is a national association of attorneys who practice or teach immigration law. AILA was founded in 1947 and currently has more than 15,000 members. The AILA membership data contains information on attorneys in wide-ranging practice settings, including law firms of various sizes, nonprofit organizations, and law schools. We geocoded the office addresses of all attorneys who were active members of AILA during fiscal year 2015 and whose practice area included removal defense. Given that many immigrants often lack resources to hire private attorneys, this study focuses on AILA attorneys who work at legal services/nonprofit organizations or law schools.
We also collected and coded information pertaining to the accredited representatives in the Executive Office for Immigration Review’s (“EOIR”) Recognition and Accreditation Program. The EOIR, under the jurisdiction of the DOJ, oversees the immigration courts. The accredited representatives in the Recognition and Accreditation Program can assist noncitizens in immigration proceedings.
We geocoded the addresses found in each record set discussed above to convert the addresses into geocoordinates. Using these geocoordinates, we produced the following measures: (1) driving distance between each detention facility and the nearest principal city within an MSA; (2)driving distance between each detention facility and the nearest nonprofit immigration attorney and the nearest EOIR accredited representative; and (3)driving distance between the facilities involved in inter-facility transfers. Of note, many geographic studies use straight-line distance measures instead of driving distance measures. This practice arose largely due to the relative ease of calculating straight-line distances, rather than due to any substantive reasons. In the Methods Appendix, we describe the basic difference between these two types of distance measures and explain why the use of driving distance is more appropriate for this study.
The third dataset we analyze consists of calls that the ICE Office of Enforcement and Removal Operations’ (“ERO”) Detention Reporting and Information Line (“DRIL”) received from detainees and community members during fiscal year 2015 that relate to specific detention facilities. Human Rights Watch obtained these records from ICE pursuant to a FOIA request. The records request stated: “HRW requests information relating to the incidence of and response to all complaints or grievances with regards to the complainant or another individual, in ICE detention facilities from fiscal year 2013 through the present.” In light of this request language, we treat the calls as a measure of facility-related grievances in this study.
The original data relating to grievances that Human Rights Watch received from ICE contained a total of 48,849 grievances pertaining to specific detention facilities. Matching these facilities to the facilities in the Detention Data and restricting the sample to only those facilities used by ICE in fiscal year 2015 produced 47,145 grievances pertaining to 304 facilities.
The DRIL initiative, launched in September of 2012, is “a toll-free service that provides a direct channel for agency stakeholders to communicate directly with ERO to answer questions and resolve concerns.” Stakeholders include “individuals in ICE custody, the public, non–governmental organizations, faith-based organizations, academic institutions, attorneys, and advocacy groups.” In the summary information accompanying its records production to Human Rights Watch, ICE noted that DRIL “does not directly investigate complaints nor does it provide outcomes for investigated complaints.” However, “[s]ome of the call concerns (complaints) are routed to field offices for review.” In addition, “[a] small number of calls are routed to the ICE Office of Professional Responsibility for assessment and possible investigation.”
It is important to note that DRIL is only one mechanism through which detainees and community members can report grievances related to detention facilities. Nonetheless, DRIL is an important source of information on facility-related grievances given that detainees may be more likely to be aware of its existence than other reporting mechanisms. For example, ICE has created an information poster for distribution to detention facilities that contains DRIL contact information for reporting concerns. A more comprehensive analysis of all grievances is difficult, if not impossible, for the reasons that we discuss later.
The Grievance Data contains information about the subject matter of the grievances, names of the facilities to which the grievances pertain, and the number of grievances per facility. ICE classified the subject matters of the grievances into the following broad categories: (1)calls related to physical and/or sexual abuse; (2)calls related to serious mental disorder or condition; (3)calls related to separation from minor child or other dependent or parental related issues; and (4)calls related to any other category. The Grievance Data contains information at the level of facilities, rather than at the level of individual detainees.
We begin our analysis by examining descriptive statistics on the detainee characteristics, detention facility characteristics, and detention outcomes. We then examine each of the three major detention outcomes of interest in this study: detention length, inter-facility transfers, and grievances. As we describe below, our analysis of detention outcomes primarily focuses on adult detainees given that juveniles are situated differently in the detention system.
We first examine various bivariate relationships, such as the relationship between confinement in a privately operated facility and detention length. We then assess whether the bivariate patterns we find are robust to the inclusion of various covariates by conducting multivariate regression analyses. For our multivariate regression analyses of detention length, we use parametric survival models based on the Weibull distribution. For the multivariate regression analyses of transfers, we use parametric survival models based on the lognormal distribution. We selected Weibull and lognormal models based on a series of tests of model fit. We fit the Weibull and lognormal models, respectively, in the accelerated failure-time metric (“AFT”) rather than in the hazard rate metric. The AFT model takes the form:
log(T) =β0 + β1x1+…+βpxp+log(ε),
in which Tis the time-to-event (the“failure time”);x1, ... ,xp are predictor variables with β regression coefficients; ε is the error term. In the detention-length analysis, the “failure” or the event of interest is obtaining release from detention. In the transfer analysis, the “failure” or the event of interest is experiencing a transfer.
For our multivariate regression analyses of grievances, we use negative binomial models. Negative binomial regression belongs to a family of generalized linear models in which the dependent variableis a count of the number of times an event occurs. The negative binomial model takes the form:
log λi = β0 + β1xi1 + … + βkxik + σεi,
in which λi is the expected count of grievances for facility i, xi1, ... , xik are the predictor variables at the facility level with β regression coefficients, and εi is the error term.
III.Key Empirical Findings
To understand who the detainees are, where they were held, and what happened to them, we examine: (1)detainee characteristics; (2)facility characteristics; and (3)detention outcomes. Appendix Table A contains a detailed description of all of the measures that we analyze below.
The ICE Detention Data contains records for 355,729 unique individuals. Who are these detainees? To address this question, we examine a number of detainee characteristics, including gender, country of citizenship, age, legal status at the time of last entry into the United States (“entry status”), and whether ICE classified the individual as an aggravated felon. Table 1 contains summary statistics on each of these characteristics for the entire detained population.
Table 1 shows that men constituted about 79% of the detainee population. In terms of the region of origin, Mexican nationals by themselves made up about 43% of the detainee population, and individuals from the Northern Triangle region of El Salvador, Guatemala, and Honduras, made up about 46% of the detainee population. Together, individuals from Mexico and Northern Triangle regions added up to 89% of the detainee population. Individuals from the rest of Latin America constituted about 5% of the detainee population, followed by individuals from Asia Pacific (3%), Africa (1%), and Europe and North America (together, 1%). Appendix Table B shows the top fifteen countries of citizenship.
Approximately 17% of the population were juveniles (under 18 years old). Adults between the ages 18 to 30 made up the largest segment of the population (42%), followed by adults between the ages 31 to 40 (26%). The population as a whole was relatively young, with mean and median ages of 28.
The ICE Detention Data contains thirty-five entry status categories. These entry status categories refer to the detainees’ legal status at their last entry into the United States. Given the relatively small sizes of many of these categories, we collapsed them into four broad categories of “asylum/refugee,” “lawful permanent resident,” “present without admission,” and “other/unknown.” Table 1 shows that a majority of the detainees were classified as present without admission (65%).
The Detention Data includes information about the detainee’s criminal history, including whether the detainee had convictions that were considered aggravated felonies. As we discussed earlier, an aggravated felony is a legal term of art in immigration law that refers to a growing host of criminal offenses that trigger removal proceedings. Only 1% of the detainee population was classified as having an aggravated felony conviction. The Detention Data also contains information about the detainee’s “most serious criminal conviction,” if any. Because our analysis suggests that ICE did not capture information on this variable for a large proportion of the detainee population, we do not further analyze this variable.
B.Detention Facility Characteristics
The Detention Data indicates that in fiscal year 2015, ICE used 638 facilities to detain individuals, including juveniles. Of the 638 facilities, 565 were used to detain adults in fiscal year 2015. Where were the 638 facilities located and what were their characteristics? We begin by considering the geographical location of the facilities. As shown in Figure 2, every state in the United States had at least one detention facility. Although not shown in Figure 2, the following U.S. territories also had at least one detention facility: Guam, Puerto Rico, the Northern Mariana Islands, and the Virgin Islands. As shown in Appendix Table C, the top five states in terms of the total number of facilities were: Texas (115 facilities), California (70 facilities), Florida (43 facilities), New York (39 facilities), and Arizona (33 facilities).
Figure 3 shows the total number of detainees held in each state in fiscal year 2015. Figure 3 underscores the relative dominance of southern and southwestern states in terms of the total detainee population. As shown in Appendix Table C, the top five states in terms of the total detainee population were: Texas (192,771 detainees), California (51,162 detainees), Arizona (44,283 detainees), Louisiana (26,481 detainees), and New Mexico (19,927 detainees). Notably, Arizona, Louisiana, and New Mexico had some of the highest levels of detainee populations in fiscal year 2015, yet they maintained relatively fewer total number of facilities compared to Texas and California.
Next, we examine a number of key detention facility characteristics, including whether a given facility is operated by a for–profit company, as well as the facility’s distance to the nearest principal city in an MSA, to the nearest nonprofit immigration attorney, and to the nearest EOIR accredited representative. We also examine the U.S. region in which the facilities are located, and facility type. In Table 2, we summarize these facility characteristics both at the facility level and at the detainee level. The estimates in the “Facility Level” column are the percentages of facilities that fall within a given category. In contrast, estimates in the “Detainee Level” column are the percentages of detainees for whom any of their detention stints fall within a given category.
Table 2 shows that while only about 10% of all facilities were operated by for-profit companies, 67% of detainees had at least one of their detention stints at a privately operated facility. Next, we examine the relative remoteness of the facilities. About 50% of all facilities were located outside of MSA principal cities, and about 64% of detainees had at least one of their detention stints in facilities located outside of MSA principal cities. At the facility level, mean distance to the nearest MSA principal city was about twenty-five miles. The range for the Distance to Nearest MSA variable was substantial: 0 to 278 miles. At the detainee level, Table 2 shows that about 50% of detainees spent at least one detention stint in a facility that was located more than thirty miles away from the nearest MSA principal city.
Another remoteness measure we examine is distance to the nearest nonprofit immigration attorney. At the facility level, mean distance to the nearest nonprofit immigration attorney was about forty-nine miles. Like the Distance to Nearest MSA variable, the range for the Distance to Nearest Nonprofit Attorney variable was also substantial: 0 to 531 miles. At the detainee level, about 58% of detainees spent at least one detention stint in a facility that was located more than thirty miles away from the nearest nonprofit immigration attorney. While the mean distance to the nearest EOIR accredited representative is relatively smaller than the mean distance to the nearest nonprofit immigration attorney, it is important to note that accredited representatives are much more limited in the type of services that they are allowed to provide.
Table 2 shows that both at the facility and at the detainee level, the facilities were heavily concentrated in the South (39% and 67%, respectively). In terms of facility type, facilities with intergovernmental service agreements (“IGSAs”) constituted the most common type of facilities (43% and 41% at the facility and detainee level, respectively). As we describe in detail in Appendix Table A, IGSAs are agreements between the federal government and a state or local government to provide detention beds in jails, prisons, and other local or state government detention facilities. These facilities are government owned, but they may be operated by either local or state agencies or by a for-profit company.
Next, we ask, what happened to the individuals held in detention? To address this question, we consider the following three types of outcomes: detention length, inter-facility transfers, and grievances. As noted earlier, our detention outcome analysis primarily focuses on adult detainees (N=296,703) given that a different set of legal requirements apply for the detention of juveniles. Before examining these three outcomes in depth, however, we first consider whether, to what extent, and on what basis individuals who had been detained in fiscal year 2015 were released from detention. In Figure 4, we show how ICE classified each detainee’s final record by examining the Release Type variable in the Detention Data. For more detailed information on each Release Type, see Appendix Table A.
As shown in Figure 4, the most common release type was Removed, with 62% of the detainees being released from detention due to removal. The next largest category was Temporarily Released, at about 20%. As described in Appendix Table A, this category includes detainees who have been released on their own recognizance, bonded out, released on supervision conditions (for example, reporting requirements) or on other alternatives to detention conditions (for example, electronic ankle monitors). The third largest category was Continued Detention, which includes all individuals who continued to be detained at the end of the fiscal year (about 12%). The last three categories of release are Paroled (about 3%), Granted Relief (about 3%), and Voluntary Departure (0.61%).
In examining detention length and transfers, wherever appropriate we conduct separate analyses for the following major release categories: Removed, Granted Relief, and Temporarily Released. For example, our analysis of detention length explores how much time various comparison groups (for instance, individuals confined in privately operated facilities versus non–privately operated facilities) spent in detention before being removed, before being granted relief, and before being temporarily released. This subgroup-analysis approach reduces the risk that our findings might be confounded by fundamental dissimilarities across individuals who experienced different types of releases.
We first examine detention length among the individuals who were released in fiscal year 2015. We noted earlier that 38,796 records pertaining to adults and juveniles are right censored in the Detention Data. For adult detainees, 35,683 records are right censored. These right-censored records pertain to individuals who continued to be detained at the end of fiscal year 2015. As we do not know how long these individuals’ time in detention lasted, we excluded them in our calculation of detention lengths shown in Figure 5. In effect, the estimates shown in Figure 5 relate to all adult detainees who were released in fiscal year 2015.
Figure 5 shows that about 17% of adult detainees who were released in fiscal year 2015 were released on the same day as their initial book-in. But many—about one-third of the adult detainees—were detained for more than thirty days. The average detention length was about thirty-eight days. Notably, the maximum value for this variable is 2,943 days, indicating that one detainee was detained for over eight years before being released in fiscal year 2015. This detainee was not alone in experiencing years of detention. For example, 1,800 adults who were released in fiscal year 2015 were detained between one to two years; another 273 adults were detained between two to three years, and 117 adults were detained more than three years, before being released in fiscal year 2015.
Next, we consider whether there are significant differences in the average detention length across the key detention facility characteristics that we discussed earlier: the type of facility operator (privately operated vs non–privately operated) and the location of the facility (within or outside of MSA principal cities). We conduct these bivariate tests separately for each of the three key release categories of interest: Removed, Granted Relief, and Temporarily Released.
To calculate the average detention length shown in Figure 6, we use the following approach: Within each major release category, we calculate the total number of detention days spent in privately operated (non–privately operated) facilities. We then divide that total by the total number of detainees who ever spent time in privately operated (non–privately operated) facilities. We follow the same approach in calculating the average detention length by facility location shown in Figure 7.
First, we note that the Granted Relief category has the highest average detention length (about sixty-eight days), followed by Temporarily Released (about fifty-five days), and Removed (about thirty-two days) (not shown in Figure 6). Second, as shown in Figure 6, there are substantial differences in the average detention length across privately operated and non–privately operated facilities. Third, as shown in Figure 7, there are also substantial differences in the average detention length across facilities that are located in MSA principal cities versus those that are located outside of MSA principal cities. Specifically, detention length is consistently and substantially higher in privately operated facilities (than in non–privately operated facilities), and in facilities located outside of MSA principal cities (than in facilities located within MSA principal cities).
To test the robustness of these findings, we conduct regression analyses of detention length that control for a variety of relevant potential confounders. These covariates include the following detainee and facility characteristics: (1)gender; (2)region of origin; (3)age at first entry into detention; (4)legal status at the time of last U.S. entry; (5)criminal history; (6)number of inter-facility transfers experienced by a detainee; (7)miles to the nearest nonprofit immigration attorney; and (8)regional location of facilities.
The regression results shown in Table 3 are based on Weibull models that we discussed earlier, and we undertake separate analyses for each major release type: Removed, Granted Relief, and Temporarily Released. The coefficients in the Weibull models we present below can be interpreted as a percent change in the expected “failure time” (in this study, detention length) with every unit increase in the independent variable, holding all other covariates in the model constant.
Our regression results are consistent with those we found in Figures 6 and 7. In each of the three release categories, being confined in privately operated facilities is associated with longer detention. For example, among those who were removed, confinement in a privately operated facility is associated with an 82% increase in days to release (100x[exp(0.60)-1] = 82.21). The same is true of confinement in a facility located outside of an MSA principal city. These patterns of results are even more pronounced for those who were granted relief and those who were temporarily released.
Some observers might ask whether these findings suggest that ICE is merely placing individuals who are likely to be detained longer into privately operated facilities, and into facilities that are located outside of major urban areas. These placement decisions may occur at the initial sorting point or at subsequent inter-facility transfers. Such a decisionmaking process would result in a pattern of detention length that would be consistent with our findings. Central to this account is the assumption that ICE is systematically and reliably predicting which detainees are likely to turn out to be long-term detainees. Our data does not allow us to formally test this possible explanation. However, at least two features of our analyses cast doubt on the validity of this account’s central assumption.
First, our findings hold across various subgroups of release categories (Removed, Granted, and Temporarily Released categories). This means that the foregoing account requires ICE to be making relatively nuanced and ongoing assessments about each detainee’s legal case—beyond mere assessments about which detainee is likely to be eventually removed, granted relief, or temporarily released. Second, our findings are robust controlling for a variety of detainee-background characteristics (see Table 3) that might confound the relationship between detention length and facility operator, and between detention length and facility location, respectively. These findings suggest that for ICE to be systematically predicting with relative accuracy which detainees will turn out to be long-term detainees, ICE must be relying on a more complex set of factors beyond these core detainee-background characteristics. Both scenarios above require us to assume a great deal about ICE’s prediction and sorting process.
Another possible explanation for the positive relationship between detention length and confinement in privately operated facilities, and confinement in facilities located outside of major urban areas, respectively, relates to the size of such facilities. Privately operated facilities tend to detain a substantially higher number of individuals on average than non–privately operated facilities. The same is true of facilities located outside of major urban areas compared to facilities located within major urban areas. Geographical concentrations of large detainee populations are likely associated with longer court backlogs, and in turn, longer time to release, all else being equal.
It is also possible that certain conditions of detention in privately operated facilities and facilities located outside of major urban areas are systematically linked to longer detention. For example, insofar as telephone calls are more expensive in privately operated facilities, or visitation is more difficult in facilities located outside of major urban areas, detainees may face greater challenges in obtaining legal counsel or obtaining materials necessary for their court hearings. If so, we might expect detainees in such facilities to seek court continuances at higher rates, leading to longer detention time. Our data do not allow us to test these possible explanations. However, our findings at a minimum call for careful future investigations into whether certain structural or situational aspects of confinement in privately operated facilities and in remote facilities might be lengthening detention.
The second detention outcome we analyze is inter-facility transfers. Two important notes about such transfers are in order. First, although transfers can occur for a variety of reasons, the Detention Data does not provide the reasons underlying the transfers. Second, ICE’s 2012 policy directive on transfers defines transfers as the movement of detainees from one Area of Responsibility (“AOR”) to another. An AOR is “[t]he geographic area of responsibility under the authority of a Field Office Director,” and it can encompass multiple states. For example, the Saint Paul Field Office’s AOR consists of Iowa, Minnesota, Nebraska, North Dakota, and South Dakota. In this study, we do not adopt this narrow definition of transfers.
Instead, we treat a movement as a transfer if ICE assigned a given record the release reason of “transferred” or “U.S. Marshals or other agency.” We refer to these movements as inter-facility transfers in this study for the following reason: ICE assigned each facility in the Detention Data a facility code that ICE has represented as uniquely identifying each facility. During our data preparation stage, we ensured that all consecutive records for any given detainee had a distinct facility code (see the Methods Appendix). In sum, all of the transfers that we analyze in this study involved movements from one location to another that had distinct facility codes.
First, we examine the prevalence of transfers. Because we cannot determine the total number of transfers for the individuals who continued to be detained at the end of fiscal year 2015, we excluded these individuals in calculating the number of transfers shown in Figure 8. As we note in Figure 8, the average number of transfers that these individuals experienced before release is a little more than one, but the range is wide: 0 to 51. Figure 8 shows that about 27% of these individuals experienced one transfer, about 15% experienced two transfers, and about 12% experienced three or more transfers. This means that about 54% of adults released in fiscal year 2015 experienced at least one transfer during their detention.
It is also illuminating to consider the type of transfers that the detainees experienced. As shown in Table 4, we coded each transfer in the Detention Data as intercity, interstate, or intercircuit. Intercity transfers refer to transfers between detention facilities that are located in different cities. Interstate transfers refer to transfers between different states. Intercircuit transfers refer to transfers between different federal judicial circuits.
All of these different types of transfers can negatively impact detainees and their legal cases. For example, transfers can sever the detainees’ familial and social contacts (in and outside the facilities where they are held), disrupt the continuity of their medical care and legal representation, and interfere with their efforts to navigate the legal system more generally. In addition to these issues, however, intercircuit transfers may impose special challenges to detainees because such transfers can mean changes in the controlling law that governs the detainees’ immigration case. Nancy Morawetz, for example, has shown that the government’s decision to transfer detainees to certain jurisdictions can “greatly increase the chances that the individual will be deported prior to any substantive review of the case.”
According to Table 4, 177,402 adults (which constitutes about 60% of 296,703 adults detained during fiscal year 2015) experienced at least one transfer. Collectively, these individuals experienced 353,704 transfers. Table 4 also shows that at the detainee level, about 86% of adult detainees who had at least one transfer experienced an intercity transfer, about 37% experienced an interstate transfer, and about 29% experienced an intercircuit transfer. At the transfer level, about 89% of all transfers were intercity transfers, about 28% were interstate transfers, and about 19% were intercircuit transfers.
Another way of examining the transfers is to consider the distance between the facilities involved. Figure 9 reports the mean and median driving distances pertaining to various transfer types. The smallest means and medians involve intercity transfers. Figure 9 shows that for intercity transfers, the mean is about 297 miles and the median is about 116 miles. The largest means and medians across different transfer types involve intercircuit transfers. Figure 9 shows that the mean for intercircuit transfers is substantial—about 820 miles. The median for intercircuit transfers is a little less than the mean—about 716 miles.
Are transfer risks associated with confinement in particular types of facilities? To address this question, we begin by examining the average number of transfers by the facility characteristics of interest in this study. Figure 10 shows the average number of transfers by the type of facility operator. Figure 11 shows the average number of transfers by the location of the facility. As before, we calculate these averages separately for each major release category of Removed, Granted Relief, and Temporarily Released.
To calculate the average number of transfers shown in Figures 10 and 11, we follow the same approach that we use in calculating the average detention lengths shown in Figures 6 and 7. For example, within each major release category, we calculate the total number of transfers out of privately operated (non–privately operated) facilities. We then divide that total by the total number of detainees who ever spent time in privately operated (non– privately operated) facilities.
Figures 10 and 11 generally indicate that transfers are more prevalent among non–privately operated facilities compared to privately operated facilities. The same is true of facilities located within MSA principal cities compared to those located outside of MSA principal cities, at least with respect to the Removed and the Temporarily Released categories. We test the robustness of these findings by estimating a set of survival models that predict the expected time to transfer, holding a number of covariates constant.
Table 5 shows that across each release category, there is a positive relationship between confinement in privately operated facilities and the expected days to transfer. Consider the results for the Removed category: Among the detainees who were eventually removed, confinement in a privately operated facility extends the expected days to transfer by about ninety-six times (exp(4.56) = 95.58), compared to being confined in a non–privately operated facility.
Likewise, Table 5 shows that across each release category, there is a positive relationship between confinement in facilities located outside of MSA principal cities and the expected days to transfer. For example, for the detainees who were eventually removed, confinement in a facility located outside of an MSA principal city extends the expected days to transfer by about four times (exp(1.32) = 3.74), compared to being confined in a facility located within an MSA principal city.
Another way to think about these results is that detainees in privately operated facilities, and in facilities located outside of MSA principal cities, respectively, are expected to experience transfers at a significantly slower rate, holding all other covariates in the model constant. The slower the failure process, the lower the risk of the expected event (in this case, transfers), all else being equal.
According to a 2009 Human Rights Watch report, “the majority of detainee transfers originate from the patchwork of local prisons and jails operating under IGSA contracts with ICE.” But that finding does not shed light on the question raised by our analysis results—namely, why does confinement in privately operated facilities, and facilities located outside of major urban areas, respectively, lower transfer risks?
The most direct evidence relevant to addressing this question is information regarding the specific reasons for the transfers. There are many different reasons why detainees may be transferred. The policy directive issued by ICE in 2012, for example, enumerates a series of officially sanctioned reasons, including reasons related to “medical or mental health care to the detainee,” the “safety and security of the detainee, other detainees, detention personnel or any ICE employee,” and the “detainee’s individual circumstances and risk factors.” Advocates and researchers have reported a variety of other reasons for transfers, including retaliation against detainees for speaking up or organizing.
To our knowledge, ICE does not maintain electronic records of why any given detainee is transferred. This critical gap in data on transfers prevents us from further analyzing the pattern of relationships that we uncovered in this study. However, one possible reason why transfer risks may be comparatively lower for privately operated facilities and facilities located outside of major urban areas might be that these facilities generally tend to have greater bed space capacities.
Such a possibility, even if empirically supported, does not favor the continued maintenance and construction of facilities with large bed space capacities. The research on the prison industrial complex has long documented the perverse incentives for incarceration generated by the expansion of prisons into communities whose local economies become reliant on the operation and profitability of those prisons. We do argue, however, that a data-driven understanding of the disparities in transfer risks across various facility types is a critical first step in developing policies to reduce the prevalence of transfers and to mitigate their resulting harm.
Finally, we examine the type and prevalence of grievances in facilities. The Grievance Data contains a total of 47,145 grievances relating to 304 of the detention facilities used by ICE in fiscal year 2015. We begin by considering the prevalence of different types of grievances. Next, we aggregate these various types of grievances to generate a total count of all grievances for each facility.
Two important caveats are in order before we begin. First, we underscore again that the grievances we analyze in this study are likely only a small fraction of all issues and incidents experienced or reported by detainees. Second, ICE’s records production to Human Rights Watch did not include definitions of the grievance types. Nor did ICE provide any explanations of how ICE classified the grievances into the various categories included in its records production. We recognize that some of the category labels used by ICE (see, for example, the “Serious or Unresolved Problem in Detention” category in Table 6) are broad and ambiguous. We also recognize that some of the categories may undercount certain kinds of grievances if those grievances were inconsistently classified. For example, grievances relating to barriers to communication with legal counsel may have been classified into the “Telephone Access” category or the “Visitation” category instead of the “Access to Counsel” category (see Table 6). For these reasons, we examine only the total count of grievances in our regression analysis.
To begin, we note that the Grievance Data contains a total of 31,417 grievances that relate to the following two categories: “Access to Legal Counsel” and “Basic Immigration Case Information.” Together, these categories concerning issues relating to access to legal representation and basic legal knowledge constitute about two-thirds of all grievances in the Grievance Data.
Table 6 contains a comprehensive list of all grievance categories as provided by ICE in its records production to Human Rights Watch. We examine the grievance statistics by type of facility operator and facility location, respectively. The statistics shown under the Type of Operation column indicate the average number of grievances (for any given grievance type) pertaining to the privately operated facilities versus those that are not privately operated. The estimates shown under the Facility Location column indicate the average number of grievances (for any given grievance type) pertaining to the facilities located within MSA principal cities versus those located outside of MSA principal cities.
For example, under the column heading, Type of Operation, Table 6 shows that privately operated facilities had an average of 2.89 grievances involving access to legal counsel. In contrast, non–privately operated facilities had an average of 0.26 grievances involving access to legal counsel. Under the column heading, Facility Location, Table 6 also shows that while facilities located within MSA principal cities had an average of 0.48 grievances involving access to legal counsel, facilities located outside of MSA principal cities had an average of 0.97 such grievances.
The overall pattern of results presented in Table 6 is similar for privately operated facilities and facilities that are located outside of MSA principal cities. In short, privately operated facilities and facilities located outside of MSA principal cities generally had a substantially higher number of grievances across almost all grievance types. In Figure 12, we show the average number of all grievances (the sum of individual grievance types) by the type of facility operator and facility location. The results are consistent with the general pattern we observed when we considered the average number of individual grievance types in Table 6. Specifically, Figure 12 shows a substantially higher overall number of grievances against privately operated facilities and facilities located outside of MSA principal cities.
To assess whether these results are robust controlling for various potential confounders, we estimate multivariate negative binomial models. In these negative binomial models, the dependent variable is the total number of grievances. The covariates in these models include: (1) Proportion Male (the proportion of detainees confined in any given facility during fiscal year 2015 who were male); (2) Proportion Mexican (the proportion of detainees confined in any given facility during fiscal year 2015 who were of Mexican origin); (3) Total Detainee Population (the total number of detainees confined in a given facility during fiscal year 2015); (4) Average Detention Stint (the average length of detention stints (in days) in a given facility in fiscal year 2015); and (5) Juvenile Facility (whether or not the facility is a juvenile facility).
The results of these negative binomial regressions are shown in Table 7. Model 2 shows that that privately operated facilities are expected to have about 175% more grievances than non–privately operated facilities, holding all other covariates in the model constant (100x[exp(1.01)-1]=174.56). Facilities located outside of MSA principal cities are expected to have about 51% more grievances than facilities located within MSA principal cities, holding all other covariates in the model constant (100x[exp(0.41)-1]
What might explain the disparities in grievance counts? If it were systematically easier to express grievances or submit grievances in privately operated facilities and facilities located outside of MSAs, grievance counts would be higher in such facilities, all else being equal. Yet a consideration of certain basic features of these facilities suggest that detainees might face equally burdensome, if not higher, barriers to filing complaints or grievances in these facilities. Research suggests that isolation—which is intrinsically exacerbated in remotely located facilities removed from community networks—can impair detainees’ cognitive and emotional functions, deprive them of their sense of agency, and prevent them from obtaining knowledge about their rights. Reports also suggest that privately operated facilities are more likely to lack transparency and public accountability. Transparency and accountability, in turn, are two basic institutional features that might be associated with open and responsive management.
The foregoing discussion focuses on whether detainees in different types of facilities may face different levels and types of challenges in submitting complaints and grievances. However, another possible explanation that requires empirical scrutiny in future research is whether and to what extent there may be differences in actual conditions of confinement that underlie disparities in grievance counts. In the criminal justice context, the DOJ’s latest review of correctional facilities in 2016 found that private contract-operated prisons had a worse record on most safety and security measures, relative to comparable facilities operated by the Federal Bureau of Prisons. This finding was important in the then-Deputy Attorney General Sally Yates’s decision to phase out private contracts. To our knowledge, no such systematic comparative assessments exist in the immigration detention context. Our findings thus provide an important starting point and foundation for future investigations on this critical issue.
Our analyses revealed a number of important patterns and significant findings. The bulk of the detained population consisted of men, and individuals from Mexico, followed by El Salvador, Guatemala, and Honduras. Together, these four countries constituted 89% of the detainee population. About 17% of the population consisted of juveniles. During fiscal year 2015, the detainees were held in a total of 638 facilities. Every state and a number of U.S. territories had at least one such facility in fiscal year 2015. While only 10% of these facilities were operated by for-profit companies, about 67% of the detainees had at least one of their detention stints at a privately operated facility.
About 50% of facilities used by ICE in fiscal year 2015 were located outside of major urban areas, and about 64% of detainees had at least one of their detention stints at such facilities. We also find that about 50% of detainees spent at least one detention stint in a facility that was located more than thirty miles away from the nearest major urban area. In addition, about 58% of detainees spent at least one detention stint in a facility that was more than thirty miles away from the nearest nonprofit immigration attorney. Taken together, these measures indicate that many detainees are confined in locations that are relatively far from community support structures and legal advocacy networks.
We also find that many detainees experienced inter-facility transfers. About 60% of adults detained during fiscal year 2015 experienced at least one transfer during their detention. About 89% of those transfers were intercity, about 28% were interstate, and about 19% involved intercircuit transfers.
These findings have important implications for not only the wellbeing of the detainees, but also for their rights to due process. For example, there is now a movement underway in many cities and states across the United States to promote greater access to legal representation for detainees. Yet such efforts are likely to be hampered in fundamental ways by the remote location of many facilities and the frequent transfer of detainees across long distances. As the Office of Inspector General has emphasized: “When ICE transfers detainees far from where they were originally detained, their legal counsel may request a release from representation because the distance and travel time or cost make representation impractical.” These challenges and disruptions not only create hardships for individual detainees, but they also have ripple effects throughout the legal system, as “[d]ifficulty arranging for counsel or accessing evidence may result in delayed court proceedings.”
Our analyses also identified notable patterns with respect to detention outcomes. We find that confinement in privately operated facilities is associated with significantly longer detention. The same is true of facilities located outside of major urban areas. Our analyses also show that privately operated facilities and facilities located outside of major urban areas have substantially higher counts of grievances. On the other hand, transfer risks are relatively lower in privately operated facilities and in facilities located outside of major urban areas.
We do not argue that the bivariate relationships that we found between the various detention outcomes and detention facility characteristics of interest (privately operated versus non–privately operated; location within MSA principal cities versus outside of MSA principal cities) are causal in nature. Whenever possible, we conducted subgroup analyses and multivariate regression analyses to control for possible confounders. Yet without knowing the assignment process of detainees to different facilities, it is difficult to draw direct causal inferences.
Nonetheless, our findings highlight the need for a careful consideration of certain basic facility characteristics in facility monitoring and reform efforts. For example, privately operated facilities and facilities located in remote areas may warrant a heightened scrutiny with respect to various detention outcomes. This is all the more so because these types of facilities traditionally have been relatively less transparent and less accountable to the public. Consider, for example, a long-running FOIA litigation in which Detention Watch Network and the Center for Constitutional Rights sought records from ICE relating to immigration detention practices. When the district court ordered the release of documents, the government chose not to appeal. However, two private prison corporations, the GEO Group and CoreCivic, intervened to stop the release of documents and filed an appeal of the court’s decision.
This study also underscores the urgent need for the government to engage in more systematic and comprehensive data collection with respect to immigration detention and to make such data widely available to researchers and the public. Importantly, any such data collection effort must be attentive to challenges that arise uniquely in detention settings. For example, underreporting of grievances is likely a serious problem given the current grievance system established by the ICE detention standards. According to a recent government report, “ICE detention standards and guidance call for informal resolution of detainee complaints through oral communication with ICE facility or contractor staff or medical personnel as appropriate.” Yet according to detainees, “staff obstructed or delayed their grievances or intimidated them, through fear of retaliation, into not complaining.” The report continued, “[t]hese deterrents may prevent detainees from filing grievances about serious concerns that should be addressed and resolved.”
Despite these challenges, investigating detention experiences and outcomes remains a critical task in light of the ongoing humanitarian and legal concerns raised by immigration detention in the United States. In a recent decision granting the habeas petition of Ravi Ragbir, a long-time New York resident and an immigrant rights activist, U.S. District Judge Katherine Forrest wrote:
[T]he process we have employed has...been unnecessarily cruel. And those who are not subjected to such measures must be shocked by it, and find it unusual. That is, that a man we have allowed to live among us for years, to build a family and participate in the life of the community, was detained, handcuffed, forcibly placed on an airplane, and today finds himself in a prison cell. All of this without any showing, or belief by ICE that there is any need to show, that he would not have left on his own if simply told to do so; there has been no showing or even intimation that he would have fled or hidden to avoid leaving as directed. And certainly there has been no showing that he has not conducted himself lawfully for years.
In requiring Ragbir’s release from immigration detention, Judge Forrest concluded: “Taking such a man, and there are many such men and women like him, and subjecting him to what is rightfully understood as no different or better than penal detention, is certainly cruel. We as a country need and must not act so. The Constitution commands better.”
A. Detention Data Preparation
To aid in our description of the Detention Data’s basic data structure, we constructed an example set of records pertaining to a single detainee (see Appendix Table D). The Initial Book-In Date is the date the detainee first came into ICE custody. The Book-In Date and the Release Date, respectively, refer to the date that the detainee was booked into and released from a given facility. The Release Type contains information about what happened to the detainee at the end of each record or detention stint. We discuss the Begin Time and End Time variables in the next Section.
We treated the records that share the same Initial Book-In Date as constituting a single “custody period.” Appendix Table D shows that the detainee had two custody periods. The first custody period consists of Records 1 through 3. The second custody period consists of Records 4 and 5. We prepared the Detention Data for analysis by taking the following steps to address any errors in ordering and issues related to duplicate or overlapping records. First, we ordered the records by sorting on the unique identification number associated with each detainee, Initial Book-In Date, Book-In Date, Release Date, Release Type, and Facility Code. The Facility Code is the unique identifier that ICE assigns to each facility.
Second, we identified and adjusted the records with overlapping information, and we removed the records that contained duplicate information. For example, if a given record contained Book-In and Release Dates that were entirely nested within the immediately preceding record, and the two records shared the same Facility Code, we dropped the nested record (the record with the smaller date range).
B. Gaps in Custody Periods and Detention Stints
In Appendix Table D, we purposely inserted gaps in dates across some of the records in order to illustrate how we addressed some of the analysis issues arising from such gaps. The example in Appendix Table D contains a gap of two days between Records 2 and 3, since Record 2 ends with a Release Date of February 7th and Record 3 begins with a Book-in Date of February 9th. A total of 20,043 such gaps exist in the Detention Data pertaining to about 5% of all detainees and less than 3% of all records. Approximately 91% of these gaps have a duration of one day.
The example in Appendix Table D also contains a gap of two days between the first and second custody periods, since the first custody period ends with a Release Date of February 9th and the second custody period begins with a Book-in Date of February 11th. A total of 772 such gaps exist in the Detention Data pertaining to 0.22% of all detainees and 0.11% of all records. The median for these gaps is five days (the range is 2 to 109 days).
For descriptive analyses involving such measures as the total number of days spent in privately operated facilities and the total number days spent in facilities located outside of MSA principal cities, we excluded these gaps from our analyses since we were unable to observe where the detainees were held during these gap periods. For the survival analyses in Table 3, we counted these gaps as continuing periods of detention in generating our time-to-event measure. Consider the Begin Time and End Time in Appendix Table D. The Begin Time refers to the number of days passed since the Initial Book-In Date at the start of each record. The End Time refers to the number of days passed since the Initial Book-In Date at the end of each record. To treat the gap periods as continuing periods of detention, we calculated the Begin Time and End Time in reference to the Initial Book-In Date of the first custody period regardless of the number of custody periods and regardless of any gaps within and across custody periods.
We tested the robustness of our survival analyses by re-estimating each of our regression models under these varying specifications: (1) excluding the gap periods; (2) excluding the detainees with multiple custody periods; and (3) excluding the gap periods and detainees with multiple custody periods. The results of these robustness checks produced substantially similar results as the results we present in Tables 3 and 5.
For any given detainee, the steps that we took to prepare the data (see the Detention Data Preparation Section in the Methods Appendix) ensured that within each custody period, consecutive records contained different Facility Codes. We then coded each record as involving a transfer if a given record’s Release Type was “transferred” or “U.S. Marshals or other agency.” In calculating the total number of transfers, we excluded each detainee’s last record. In effect, we did not treat the last record as involving a transfer even if that record’s Release Type was “transferred” or “U.S. Marshals or other agency,” since in such instances we could not calculate the distance between the originating and receiving facilities.
Although ICE has represented that the Facility Code uniquely identifies each facility, we found sixty-four records pertaining to sixty-one detainees in which two consecutive records in a custody period differed in their Facility Codes but shared the same address. We suspect that these transfers involved movements from one part of a building complex with one facility code, to another part of the same building complex with a different facility code.
D. Coding Facilities as Privately Operated
We classified detention facilities as privately operated using the following steps. First, we used a list of facilities compiled by the Enforcement and Removal Operations’ Custody Management Division. This list contains information about 208 facilities that were in use during fiscal year 2015, including an indicator of whether any given facility was operated by a private for-profit company (for example, GEO Group), or local government (for example, a county or city sheriff’s department).
Second, we reviewed a set of contracts available in ICE’s FOIA library that pertain to agreements between ICE and private for-profit companies. We then used the Wayback Machine to examine archived versions of the companies’ websites to compile a list of all facilities that the companies operated in fiscal year 2015. Although we recognize that a facility that was privately operated in fiscal year 2015 may not have been privately operated in earlier fiscal years, our coding is time-invariant. A total of 46,202 records (6% of records) pertaining to 23,492 detainees (3% of detainees) in the Detention Data pre-date fiscal year 2015.
E. Distance Measures
Rather than use a straight-line measure of distance based on changes in latitude and longitude coordinates, we used the distance generated by Google Maps based on driving routes between various points of interest. The advantage of using the driving distance is that by accounting for the existence or nonexistence of roads, freeways, interstates, mountains, bodies of water, and other geographical features of land, we better approximate the actual distance that legal counsel, kin, or advocates would have to travel to reach a given facility.
The use of driving distance, however, poses challenges for measuring distances involving points of interest that are surrounded by seas and oceans. While we can measure the driving distance to the nearest MSA principal city in Oahu, Hawaii and in Puerto Rico because those two islands contain MSA principal cities, we cannot do the same for other U.S. territories and Sitka, Alaska, which do not have MSAs. We thus treat the facilities in latter locations (N=7) as missing on Distance to Nearest MSA. The same problem exists in terms measuring the driving distance to the nearest nonprofit immigration attorney, and the driving distance to the nearest accredited representative. In locations like Sitka, Alaska, where such attorneys or representatives do not exist, crossing a body of water is necessary to reach the nearest attorney or representative. We treat the facilities in these types of locations as missing on Distance to Nearest Nonprofit Attorney (N=16), and Distance to Nearest Accredited Representative (N=14).
To estimate the distance between facilities involved in inter-facility transfers, however, we implemented a different strategy. Because distance measures for inter-facility transfers represent actual distances that detainees were required to travel (whether by air or ground transportation), excluding facilities that are located in territories, Hawaii, and Sitka, Alaska, would result in a significant underestimation of these distances. Thus, we rely on the driving distance wherever driving between a given pair of facilities is possible, and on the Haversine formula (a straight-line distance measure) wherever a transfer necessitates crossing a body of water.
F. Geocoding Facilities
To obtain latitude and longitude coordinates for the facilities in the Detention Data, we used a multi-pronged approach. First, we matched the facilities in the Detention Data to the facilities in the National Immigrant Justice Center’s ICE detention facility list (“NIJC Data”), which the NIJC obtained from ICE through a FOIA request. The NIJC Data contains addresses for the facilities included in its list. The matching process allowed us to assign addresses to 521 of the 679 facilities in the Detention Data (77%). We then used the geocode function of R’s ggmap package to assign geocoordinates to each of the 521 facilities.
For the remaining 158 facilities in the Detention Data, we found the city and state in which each facility is located using information included in the TRAC detention facility reports. We then queried Google Maps using the facility name and its city and state. As shown in Appendix Table E (see Google Maps category), we were able to assign geocoordinates to eighty-one additional facilities in the Detention Data (12%) using this method.
To assign geocoordinates for the rest of the facilities in the Detention Data, Google Maps required additional information beyond facility name, city, and state. We compiled the addresses for most of these facilities using the following sources: (1) the yellow page directories; (2) the websites of the non–profit and for–profit organizations that operated the facilities; (3) the websites of governmental entities, including the Bureau of Prisons, U.S. Marshals Service, and county and city correctional departments; (4) the Global Detention Project website; and (5) the websites of various third party agencies describing or evaluating the facilities. We were able to assign geocoordinates to another 10% of the facilities in the Detention Data using the facility addresses we compiled from these sources (see Yellow Pages, Operating Organization Websites, Government Websites, Global Detention Project Website, and Other Sources categories).
Finally, for the remaining 3% of facilities in the Detention Data for which we could not assign geocoordinates using the above methods, we assigned them geocoordinates that corresponded to the city and state where the facility is located (see City and State Information (TRAC) category). All of these facilities were one of several branches operated by the same non–profit organizations located within same cities.
[*] *.. This research was supported by the Carnegie Corporation of New York and the California Wellness Foundation. The statements made and views expressed are solely the responsibility of the authors. We are grateful to Ben Johnson, Sue Long, Grace Meng, Brian Root, and Tizita Wasihun. We thank Tendayi Achiume, Sameer Ashar, Lenni Benson, Guillermo Cantor, Jennifer Chacón, Ingrid Eagly, Annie Lai, and Jaya Ramji-Nogales for their insightful comments. Kris Coombs, Joanna Cortez Hernández, Danielle Flores, Hilary Olson, and Tiffany Panganiban provided excellent research assistance. We thank the participants of the Criminalizing Immigrants Conference at Cornell University, the UCLA Center for the Study of International Migration Seminar Series, and the Berkeley International Migration and Refugee Law Workshop for providing helpful feedback. Direct all correspondence to Emily Ryo, USC Gould School of Law, 699 Exposition Blvd., Los Angeles, CA 90089. Email: email@example.com.
[†] †.. Associate Professor of Law and Sociology, University of Southern California Gould School of Law.
[‡] ‡.. Ph.D. Candidate in Sociology, University of California, Los Angeles.
. Zadvydas v. Davis, 533 U.S. 678, 690 (2001).
. McMillan v. Pennsylvania, 477 U.S. 79, 98 n.2 (1986) (“The combination of stigma and loss of liberty involved in a conditional or absolute sentence of imprisonment sets that sanction apart from anything else the law imposes.”) (internal citation omitted); Vitek v. Jones, 445 U.S. 480, 492 (1980) (“The loss of liberty produced by an involuntary commitment is more than a loss of freedom from confinement.”); Castle v. United States, 399 F.2d 642, 651 n.16 (5th Cir. 1968) (“[I]mprisonment brings not only a financial loss in wages but also the loss of all other advantages and privileges of being at liberty.”).
. We define immigration detention as a confinement system based on the federal government’s power to hold individuals pending their immigration proceedings. By contrast, criminal incarceration refers to a confinement system based on state or federal government’s authority to hold individuals charged with, or convicted of, a criminal offense. See Dora Schriro, U.S. Dep’t of Homeland Sec., Immigration Detention Overview and Recommendations 4 (2009), https://www.ice.gov/doclib
/about/offices/odpp/pdf/ice-detention-rpt.pdf [https://perma.cc/FC6L-WSH8%5D; Office of the Fed. Det. Tr., Detention Needs Assessment and Baseline Report 4 n.2 (2001), https://www.justice.gov
/archive/ofdt/federal_detention_report_2002.pdf [https://perma.cc/MSW7-2Z2C%5D. Some observers, however, use the term immigration incarceration to refer to immigration detention, in part to highlight the punitive nature and severe consequences of immigration detention. See, e.g., Semuteh Freeman & Lauren Major, Immigration Incarceration: The Expansion and Failed Reform of Immigration Detention in Essex County, NJ (2012), http://www.law.nyu.edu/sites/default/files
/upload_documents/Immigration%20Incarceration.pdf [https://perma.cc/56BD-DXVW%5D; César Cuauhtémoc García Hernández, Naturalizing Immigration Imprisonment, 103 Calif. L. Rev. 1449 (2015); Raha Jorjani, Locked Up: Criminal and Immigration Incarceration in America, Keynote Address (Mar. 16, 2010) in 4 DePaul J. for Soc. Just. 1 (2010).
. Dora B. Schriro, Improving Conditions of Confinement for Immigrant Detainees: Guideposts Toward a Civil System of Civil Detention, in The New Deportations Delirium: Interdisciplinary Responses 57, 66 (Daniel Kanstroom & M. Brinton Lykes eds., 2015).
. See Exec. Order No. 13767, 82 Fed. Reg. 8793 (Jan. 25, 2017).
. Memorandum from John Kelly, Sec’y of Homeland Sec., to Kevin McAleenan, Acting Comm’r of U.S. Customs & Border Protection et al. 8–9 (Feb. 20, 2017), https://www.dhs.gov/sites
/default/files/publications/17_0220_S1_Implementing-the-Presidents-Border-Security-Immigration-Enforcement-Improvement-Policies.pdf [https://perma.cc/LVX3-EGMY%5D; Immigration and Customs Enforcement & Customs and Border Protection FY18 Budget Request: Hearing Before the Subcomm. on Homeland Sec. of the H. Comm. on Appropriations, 115th Cong. (2017) (statement of Thomas Homan, Acting Director, U.S. Immigration and Customs Enforcement), https://www.dhs.gov/news/2017/06/13
. Caitlin Dickerson, Plan Would Limit Protections for Immigrants Held in Jails, N.Y. Times, Apr. 14, 2017, at A1.
. See generally, e.g., Am. Civil Liberties Union et al., Fatal Neglect: How ICE Ignores Deaths in Detention (2016), https://www.detentionwatchnetwork.org/sites/default/files/reports/Fatal
%20Neglect%20ACLU-DWN-NIJC.pdf [https://perma.cc/J4VP-NU75]; S. Poverty Law Ctr. et al., Shadow Prisons: Immigrant Detention in the South (2016), https://www.splcenter.org/sites
/default/files/ijp_shadow_prisons_immigrant_detention_report.pdf [https://perma.cc/2GMD-M9RD%5D; Human Rights Watch & Cmty. Initiatives for Visiting Immigrants in Confinement, Systemic Indifference: Dangerous & Substandard Medical Care in US Immigration Detention (2017), https://www.hrw.org/sites/default/files/report_pdf/usimmigration0517_web_0.pdf [http://perma.cc
/VQW5-UNMV]; Pa. State Law Ctr. for Immigrants’ Rights Clinic, Imprisoned Justice: Inside Two Georgia Immigrant Detention Centers (2017), https://projectsouth.org/wp-content/uploads
. For notable exceptions, see generally Ingrid V. Eagly & Steven Shafer, A National Study of Access to Counsel in Immigration Court, 164 U. Pa. L. Rev. 1 (2015) (using data on immigration courts to examine access to counsel in immigration proceedings, including for detained immigrants); Ingrid Eagly et al., Detaining Families: A Study of Asylum Adjudication in Family Detention, 106 Calif. L. Rev. 785 (2018) (using data on immigration courts to examine family detention). For challenges that researchers face in obtaining government data related to immigration detention, see, for example, Transactional Records Access Clearinghouse, Huge Increase in Transfers of ICE Detainees (2009), http://trac.syr.edu/immigration/reports/220/#20 [https://perma.cc/3NXM-2BTB%5D; Donald Kerwin et al., Piecing Together the US Immigrant Detention Puzzle One Night at a Time: An Analysis of All Persons in DHS-ICE Custody on September 22, 2012, 3 J. on Migration & Hum. Security 330, 331–32 (2015).
. See generally Jennifer M. Chacón, Privatized Immigration Enforcement, 52 Harv. C.R.-C.L. L. Rev. 1 (2017) (evaluating privatization in immigration law).
. For additional recent discussions on the historical and political context of immigration detention, see Emily Ryo, Fostering Legal Cynicism Through Immigration Detention, 90 S. Cal. L. Rev. 999 (2017).
. Antiterrorism and Effective Death Penalty Act of 1996, Pub. L. No. 104-132, § 440(a), 110 Stat. 1214 (1996).
. Illegal Immigration Reform and Immigrant Responsibility Act of 1996, Pub. L. No. 104-208, § 133, 110 Stat. 3009 (1996).
. César Cuauhtémoc García Hernández, Immigration Detention as Punishment, 61 UCLA L. Rev. 1346, 1370 (2014); Margaret H. Taylor, The 1996 Immigration Act: Detention and Related Issues, 74 Interpreter Releases 209, 216 (1997).
. García Hernández, supra note 18, at 1370–71.
. Nancy Morawetz, Understanding the Impact of the 1996 Deportation Laws and the Limited Scope of Proposed Reforms, 113 Harv. L. Rev. 1936, 1939 (2000).
. There is some debate as to whether this growth can be partly attributed to what is commonly known as the “detention bed quota” or the “detention bed mandate.” Since 2009, Congress has mandated that the DHS maintain a certain minimum number of detention beds on a daily basis. The quota in 2009 was 33,400 beds, which increased to 34,000 in 2011. See Patrisia Macías-Rojas, From Deportation to Prison: The Politics of Immigration Enforcement in Post–Civil Rights America 73–74 (2016).
. Exec. Order No. 13768, 82 Fed. Reg. 8799 (Jan. 25, 2017); Exec. Order No. 13767, 82 Fed. Reg. 8793 (Jan. 25, 2017); Memorandum from John Kelly, supra note 7.
. Between February and September of 2016, non–criminal removals were 41% of all removals. That figure rose to 43% during the same time period in 2017. See U.S. Immigration & Customs Enf’t, Fiscal Year 2017 ICE Enforcement and Removal Operations Report (Dec. 13, 2017), https://www.ice.gov/removal-statistics/2017 (select “Local Statistics” tab; then view PDF of “Local Statistics 2013–2016” for the 2016 statistics, and view PDF of “Local Statistics 2017” for the 2017 statistics).
. Exec. Order No. 13767, 82 Fed. Reg. 8793 (Jan. 25, 2017); see also Exec. Order No. 13768, 82 Fed. Reg. 8799 (Jan. 25, 2017); Memorandum from John Kelly, supra note 7.
. Regarding the Fiscal Year 2018 President’s Budget Request: Hearing Before the Subcomm. on Homeland Sec. of the H. Comm. on Appropriations, 115th Cong. 3–4 (2017) (statement of Thomas Homan, Acting Director, U.S. Immigration and Customs Enforcement), https://docs.house.gov/meetings
. U.S. Dep’t of Homeland Sec., supra note 5, at 11; U.S. Dep’t of Homeland Sec., CBP Border Security Report: Fiscal Year 2017 1 (2017), https://www.cbp.gov/sites/default/files
/assets/documents/2017-Dec/cbp-border-security-report-fy2017.pdf [https://perma.cc/ERP5-L5PE%5D; see also Nick Miroff, Arrests Along Mexico Border Drop Sharply Under Trump, New Statistics Show, Wash. Post (Dec. 5, 2017), http://wapo.st/2Atmy6j?tid=ss_tw&utm_term=.53a99e255c63 [http://perma.cc
/9ST2-2PHD] (explaining that border apprehensions “show a sharp drop . . . immediately after President Trump’s election win, possibly reflecting the deterrent effect of his rhetoric on would-be border crossers; starting in May, the number of people taken into custody began increasing again”).
. U.S. Dep’t of Homeland Sec., supra note 5, at 10 fig. 10.
. García Hernández, supra note 18, at 1370–71; Ryo, supra note 15, at 1024–34; Dora Schriro, Improving Conditions of Confinement for Criminal Inmates and Immigrant Detainees, 47 Am. Crim. L. Rev. 1441, 1442 (2010).
. Rodriguez v. Robbins, 804 F.3d 1060, 1065 (9th Cir. 2015), rev’d sub nom. Jennings v. Rodriguez, 138 S. Ct. 830 (2018).
. Am. Immigration Council, Two Systems of Justice: How the Immigration System Falls Short of American Ideals of Justice 1 (2013), https://www.americanimmigrationcouncil.org
/sites/default/files/research/aic_twosystemsofjustice.pdf [https://perma.cc/L4HC-6JYR%5D; David Cole, In Aid of Removal: Due Process Limits on Immigration Detention, 51 Emory L.J. 1003, 1006–08 (2002); Philip L. Torrey, Rethinking Immigration’s Mandatory Detention Regime: Politics, Profit, and the Meaning of “Custody”, 48 U. Mich. J.L. Reform 879, 880–81 (2015).
. For a helpful and more detailed discussion on the statutory framework and the categories of noncitizens subject to detention, see Powers of the U.S. Government to Detain Noncitizens, 8 Immigr. L. & Proc. (MB) §108.02 (2017).
. Immigration and Nationality Act § 236(a), 8 U.S.C. § 1226(a) (2018).
. Immigration and Nationality Act § 236(a)(2); see also Ortega-Cervantes v. Gonzales, 501 F.3d 1111, 1115 (9th Cir. 2007) (noting that “the phrase ‘release on recognizance’” is used “as another name for ‘conditional parole’ under § 1226(a)”).
. Immigration and Nationality Act § 236(b); see also 8 C.F.R. § 236.1(c)(9) (2018).
. See generally Powers of the U.S. Government to Release Detained Noncitizens, 8 Immigr. L. & Proc. (MB) §108.03 (2018) (discussing mandatory detention).
. Immigration and Nationality Act §§ 235(b)(2)(A), 236(c), 236A(a), 241(a)(2).
. 8 C.F.R. § 1.2 (2018). Arriving aliens subject to mandatory detention include individuals seeking asylum pending a final determination of whether they demonstrate a credible fear of persecution. Immigration and Nationality Act § 235(b)(1)(B)(iii)(IV).
. See Legal Challenges to Detention, 8 Immigr. L. & Proc. (MB) §108.05 (2018). See generally Farrin R. Anello, Due Process and Temporal Limits on Mandatory Immigration Detention, 65 Hastings L.J. 363 (2014) (discussing federal courts’ treatment of prolonged detention); Philip L. Torrey, Jennings v. Rodriguez and the Future of Immigration Detention, 20 Harv. Latinx L. Rev. 171 (2017) (same).
. Immigration and Nationality Act § 241(g).
. Avramenkov v. Immigration & Naturalization Serv., 99 F. Supp. 2d 210, 213 (D. Conn. 2000); Comm. of Cent. Am. Refugees v. Immigration & Naturalization Serv., 682 F. Supp. 1055, 1064 (N.D. Cal. 1988) (noting that “[t]he Attorney General has the authority, conferred by statute, to choose the location for detention and to transfer aliens to that location. . . . Such a transfer, standing alone, does not constitute a violation of plaintiffs’ due process or statutory rights”); see also Rady v. Ashcroft, 193 F. Supp. 2d. 454, 456–57 (D. Conn. 2002) (same); Sasso v. Milhollan, 735 F. Supp. 1045, 1048 (S.D. Fla. 1990) (same). But see César Cuauhtémoc García Hernández, Due Process and Immigrant Detainee Prison Transfers: Moving LPRs to Isolated Prisons Violates Their Right to Counsel, 21 Berkeley La Raza L.J. 17, 17 (2011) (arguing that the transfer of lawful permanent residents from one detention facility to another “frequently violates the Fifth Amendment’s due process right to counsel for lawful permanent residents”).
. See Human Rights Watch, Locked Up Far Away: The Transfer of Immigrants to Remote Detention Centers in the United States (2009), https://www.hrw.org/sites/default/files
/reports/us1209webwcover.pdf [https://perma.cc/J2WB-LXX3]. For an updated Human Rights Watch report on transfers, see Human Rights Watch, supra note 11.
. See Letter from Phyllis A. Coven, Acting Dir., U.S. Immigration & Customs Enf’t, to Alison Parker, Deputy Dir., U.S. Program, Human Rights Watch (Feb. 22, 2010), https://www.hrw.org/sites
/default/files/related_material/Coven%20to%20HRW%20Feb%202010.pdf [https://perma.cc/LF2N-KNFF%5D. In 2011, ICE revised the transfer provisions in its Performance-Based National Detention Standards. See U.S. Immigration & Customs Enf’t, Performance-Based National Detention Standards 2011, at 443–50 (2013), https://www.ice.gov/doclib/detention-standards/2011/pbnds2011
. Maria Baldini-Potermin, § 3:3 ICE Office of Chief Counsel, in Immigration Trial Handbook (2017).
. U.S. Dep’t of Homeland Sec., supra note 13, at 2.
. Id. at 2, 11 n.14; see also Matt Zapotosky, Justice Department Will Again Use Private Prisons, Wash. Post, Feb. 24, 2017, at A4.
. U.S. Comm’n on Civil Rights, supra note 56, at 25.
. Office of Inspector Gen., U.S. Dep’t of Homeland Sec., supra note 56, at 3.
. See generally, e.g., Cole, supra note 33 (discussing constitutional limitations on immigration detention); Alina Das, Immigration Detention: Information Gaps and Institutional Barriers to Reform, 80 U. Chi. L. Rev. 137 (2013) (discussing policy issues related to immigration detention); García Hernández, supra note 3 (examining the legal and political decisions that affect immigration detention); Stephen H. Legomsky, The Detention of Aliens: Theories, Rules, and Discretion, 30 U. Miami Inter-Am. L. Rev. 531 (1999) (discussing detention decisions); Margaret H. Taylor, Dangerous by Decree: Detention Without Bond in Immigration Proceedings, 50 Loy. L. Rev. 149 (2004) (examining the executive branch’s legal authority for immigration detention without bond).
. See generally, e.g., Robert Barde & Gustavo J. Bobonis, Detention at Angel Island: First Empirical Evidence, 30 Soc. Sci. Hist. 103 (2006) (analyzing immigration detention at San Francisco’s Angel Immigration Station from 1913 to 1919); Kalina Brabeck & Qingwen Xu, The Impact of Detention and Deportation on Latino Immigrant Children and Families: A Quantitative Exploration, 32 Hisp. J. Behav. Sci. 341 (2010) (analyzing immigration detention’s impact on immigrant families); Nancy Hiemstra & Deirdre Conlon, Beyond Privatization: Bureaucratization and the Spatialities of Immigration Detention Expansion, 5 Territory, Pol., Governance 252 (2017) (exploring the role of beauracratization in immigration detention’s expansion); Luis H. Zayas et al., The Distress of Citizen-Children with Detained and Deported Parents, 24 J. Child & Fam. Stud. 3213 (2015) (exploring the impact of immigration detention on immigrant children).
. See generally, e.g., Ingrid V. Eagly, Remote Adjudication in Immigration, 109 Nw. L. Rev. 933 (2015) (examining the role of remote adjudication in immigration cases); Kerwin et al., supra note 10 (analyzing nationwide detention data for a single night and arguing the federal government should work to provide this data more broadly and consistently); Emily Ryo, Legal Attitudes of Immigrant Detainees, 51 Law & Soc’y Rev. 99 (2017) (examining immigrant detainees’ legal attitudes); Emily Ryo, Representing Immigrants: The Role of Lawyers in Immigration Bond Hearings, 52 Law & Soc’y Rev. 503 (2018) (investigating the role of lawyers in immigration bond hearings); Emily Ryo, Predicting Danger in Immigration Courts, Law & Soc. Inquiry (forthcoming 2018) (analyzing immigration judges’ decisions on dangerousness in bond hearings).
. See, e.g., Hiemstra & Conlon, supra note 61; Torrey, supra note 33.
. Chacón, supra note 12, at 31–32.
. Denise Gilman & Luis A. Romero, Immigration Detention, Inc. 6 J. on Migration & Hum. Security 145 passim (2018).
. Alison Mountz, Mapping Remote Detention: Dis/location through Isolation, in Beyond Walls and Cages: Prisons, Borders, and Global Crisis 91, 99–100 (Jenna M. Lloyd et al. eds., 2012).
. Lauren L. Martin, ‘Catch and Remove’: Detention, Deterrence, and Discipline in US Noncitizen Family Detention Practice, 17 Geopolitics 312, 326 (2012). Of note, studies have found that legal service organizations are one such supporting institution whose involvement is consistently related to more favorable outcomes in custody redetermination hearings—an essential release mechanism from detention. Eagly & Shafer, supra note 10; Emily Ryo, Detained: A Study of Immigration Bond Hearings, 50 Law & Soc’y Rev. 117 (2016); Ryo, Representing Immigrants, supra note 62.
. Fiscal year 2015 began on October 1, 2014 and ended on September 30, 2015. The Detention Data, however, contains information on sixty-one detainees who were released on October 13, 2015, indicating that the last date of the data extraction by ICE was October 13, 2015.
. Although TRAC obtained similar data from ICE pertaining to fiscal years 2013, 2014, and 2016, ICE produced these data in such a way that they are not amenable to meaningful analysis. For example, some of the data are missing a large set of records, while others contain a large set of seemingly duplicate records that cannot be identified accurately as duplicates.
. As discussed earlier, some of the records in the Detention Data predate fiscal year 2015; 679 facilities pertain to all of the records in the Detention Data, including the records that predate fiscal year 2015.
. We used the U.S. Office of Management and Budget’s (“OMB”) February 2013 delineation of metropolitan statistical areas (“MSAs”) and principal cities for our coding purposes, as doing so allowed us to identify the location of each facility in the Detention Data with respect to the principal cities at the beginning of fiscal year 2015. See id.
. See Exec. Office for Immigration Review, Recognition & Accreditation (R&A) Program, U.S. Dep’t of Justice, https://www.justice.gov/eoir/recognition-and-accreditation-program [https://perma
.cc/7YQ6-T467] (last updated Oct. 5, 2018). We used the list that appears on the archived Executive Office for Immigration Review (“EOIR”) website, which was updated on September 7, 2015. Exec. Office for Immigration Review, Accredited Representatives Roster, U.S. Dep’t of Justice (Sept. 7, 2015), https://www.justice.gov/sites/default/files/pages/attachments/2015/09/08/raroster-reps.pdf [https:
. See generally Francis P. Boscoe et al., A Nationwide Comparison of Driving Distance Versus Straight-Line Distance to Hospitals, 64 Prof. Geographer 188 (2012).
. In a note accompanying the records production, ICE indicated that records without detention facility names are calls that did not directly pertain to detention facilities. We did not include these calls in our analysis.
. See U.S. Gov’t Accountability Office, Immigration Detention: Additional Actions Needed to Strengthen Management and Oversight of Detainee Medical Care, GAO-16-231, at 11 (2016), https://www.gao.gov/assets/680/675484.pdf [https://perma.cc/6VN6-BS5Z].
. Cf. Eagly & Shafer, supra note 10, at 89 (removing juvenile cases in analyzing case outcomes and discussing other researchers’ decisions to do the same).
. To determine which possible functional forms best fit our data, we estimated the empirical Nelson-Aalen cumulative hazard function and plotted it against the cumulative hazard of the Cox-Snell residuals, which should be a straight 45o line if the model fit the data well. We also examined the Akaike information criterion and the Bayesian information criterion, two conventional information-based criteria that assess model fit. We compared the results of these two diagnostic assessments across five different types of survival models, Cox, Weibull, exponential, lognormal, and loglogistic, and found that Weibull and lognormal models best fit our data. See Mario Cleves et al., An Introduction to Survival Analysis Using Stata 221–25 (3d ed. 2016).
. Negative binomial regression is a generalization of Poisson regression with an extra parameter to model overdispersion, which occurs when the conditional variance exceeds the conditional mean. See Joseph M. Hilbe, Negative Binomial Regression 1–5 (2d ed. 2011).
. We generally followed the United Nations’ classification system to sort the countries of citizenship in the Detention Data into major world regions. See United Nations, Methodology: Standard Country or Area Codes for Statistical Use (M49), https://unstats.un.org/unsd
. About 63% of detainees who were designated as having an aggravated felony were missing on the “most serious criminal conviction” variable.
. As we noted earlier, the Detention Data is longitudinal and contains information about the individual detention stints of all individuals detained in fiscal year 2015. Some of these detention stints occurred before fiscal year 2015 and in facilities that were not used by ICE in fiscal year 2015.
. Of the facilities in U.S. territories, seven were in Puerto Rico, three in Guam, two in the Virgin Islands, and one in Northern Mariana Islands.
. We present the detainee-level statistics in this way because detainees can move in and out of various facilities over the course of their detention.
. See generally William Wilberforce Trafficking Victims Protection Reauthorization Act of 2008, Pub. L. No. 110-457, 122 Stat. 5044; Stipulated Settlement Agreement, Flores v. Reno, No. 85-4544-RKJ (Px) (C.D. Cal. Jan. 17, 1997). The Detention Data does not contain information about whether the juveniles are unaccompanied or accompanied minors.
. For robustness checks on the results presented in Table 3 under varying specifications, see the Methods Appendix.
. U.S. Immigration & Customs Enf’t, supra note 47.
. As we explain in detail in the Methods Appendix, however, a small number of transfers (N=64) involved a movement from one location to another that had distinct facility codes yet shared the same address.
. See, e.g., Inter-Am. Comm’n on Human Rights, Report on Immigration in the United States: Detention and Due Process 26–27, 137–40 (2010); García Hernández, supra note 43, at 17; Steering Comm. of the N.Y. Immigrant Representation Study Report, Accessing Justice: The Availability and Adequacy of Counsel in Removal Proceedings, 33 Cardozo L. Rev. 357, 363 (2011); Mark Noferi, Cascading Constitutional Deprivation: The Right to Appointed Counsel for Mandatorily Detained Immigrants Pending Removal Proceedings, 18 Mich. J. Race & L. 63, 76–77 (2012).
. Peter L. Markowitz & Lindsay C. Nash, Constitutional Venue, 66 Fla. L. Rev. 1153, 1200 (2015) (noting that “federal immigration authorities’ unchecked control over venue allows them to manipulate the controlling law of a case because the case will be governed by the law of the federal circuit in the jurisdiction in which the immigration court sits”).
. Nancy Morawetz, Detention Decisions and Access to Habeas Corpus for Immigrants Facing Deportation, 25 B.C. Third World L.J. 13, 16 (2005).
. For robustness checks on the results presented in Table 5 under varying specifications, see the Methods Appendix.
. Human Rights Watch, supra note 45, at 21; see also Human Rights Watch, supra note 11, at 24 (“More than half of all transfers involved a facility that has an Intergovernmental Service Agreement with ICE to hold immigration detainees. These facilities are most commonly state or local criminal jails and prisons . . . .”).
. U.S. Immigration & Customs Enf’t, supra note 47, at 3.
. Nancy Hiemstra, “You Don’t Even Know Where You Are”: Chaotic Geographies of U.S. Migrant Detention and Deportation, in Carceral Spaces: Mobility and Agency in Imprisonment and Migrant Detention 68 (Dominique Moran et al. eds., 2013); D. Conlon & N. Hiemstra, Examining the Everyday Micro-Economies of Migrant Detention in the United States, 69 Geographica Helvetica 335, 342 (2014).
. On September 15, 2017, the first author filed a records request under FOIA with ICE seeking all records that reflect or describe policies, procedures, or systems used to make placement and transfer determinations. On May 10, 2018, ICE produced a one-page letter containing three ICE websites pertaining to ICE’s official policies on transfers. We thus continue to lack information about how ICE makes these determinations in practice.
. Our analysis of the ICE detention-facility dataset that contains information on the facilities’ bed space capacity generally supports this conclusion. See 2015 ICE Detention Facility Listing, U.S. Immigration & Customs Enf’t (Oct. 3, 2017), https://www.ice.gov/foia/library (expand “Detention Facility Statistics;” then follow “2015 ICE Detention Facility Listing” hyperlink); see also The GEO Grp., Inc., 2011 Annual Report 2 (highlighting the large bed-space capacity of its facilities as its competitive market advantage); Corrections Corporation of America, 2009 Annual Report on Form 10-K 9 (2009) (same).
. See generally, e.g., Rose M. Brewer & Nancy A. Heitzeg, The Racialization of Crime and Punishment: Criminal Justice, Color-Blind Racism, and the Political Economy of the Prison Industrial Complex, 51 Am. Behav. Scientist 625 (2008); Heather Ann Thompson, Why Mass Incarceration Matters: Rethinking Crisis, Decline, and Transformation in Postwar American History, 97 J. Am. Hist. 703 (2010); Rebecca U. Thorpe, Perverse Politics: The Persistence of Mass Imprisonment in the Twenty-First Century, 13 Persp. on Pol. 618 (2015).
. Mountz, supra note 66, at 95–96 (discussing the ways in which isolation deprives detainees of essential knowledge and information); Alison Mountz et al., Conceptualizing Detention: Mobility, Containment, Bordering, and Exclusion, 37 Progress in Hum. Geography 522, 529 (2013) (discussing research that documents the negative psychological impacts of isolation on detainees).
. See Am. Civil Liberties Union of N.M., Outsourcing Responsibility: The Human Cost of Privatized Immigration Detention in Otero County 59 (2011), https://www.aclu-nm.org/sites
/default/files/wp-content/uploads/2011/01/OCPC-Report.pdf [https://perma.cc/2MPC-P4FL]; U.S. Dep’t of Homeland Sec., supra note 13, at 45.
. See Chacón, supra note 12, at 21–32.
. See Office of the Inspector Gen., U.S. Dep’t of Justice, Review of the Federal Bureau of Prisons’ Monitoring of Contract Prisons, at i (2016), https://oig.justice.gov/reports
. Memorandum from Sally Q. Yates, supra note 52. Attorney General Jeff Sessions, nominated by President Trump, subsequently rescinded the Memorandum from Deputy Attorney General Yates, noting that “[t]he memorandum changed long-standing policy and practice, and impaired the Bureau’s ability to meet the future needs of the federal correctional system.” Memorandum from Jefferson B. Sessions III, Attorney Gen., to the Acting Director, Fed. Bureau of Prisons (Feb. 21, 2017), https://www.bop.gov/resources/news/pdfs/20170224_doj_memo.pdf [https://perma.cc/4CNQ-G4TY%5D.
. See Chacón, supra note 12, at 25 (concluding that “systematic comparative evidence is lacking” with respect to conditions of confinement in immigration detention).
. See generally, e.g., Cal. Coal. for Universal Representation, California’s Due Process Crisis: Access to Legal Counsel for Detained Immigrants (2016), https://www.nilc.org
/X8LN-XSTB]; Andrea Black & Joan Friedland, Nat’l Immigration Law Ctr., Blazing a Trail: The Fight for Right to Counsel in Detention and Beyond (Richard Irwin ed., 2016), https://www.nilc.org/wp-content/uploads/2016/04/Right-to-Counsel-Blazing-a-Trail-2016-03.pdf [https://perma.cc/JQJ3-ACPC%5D; Maggie Corser, Ctr. for Popular Democracy, Access to Justice: Ensuring Counsel for Immigrants Facing Deportation in the D.C. Metropolitan Area (Emily Tucker et al. eds., 2017), https://populardemocracy.org/sites/default/files/DC_Access_to_Counsel_rev4
. Office of Inspector Gen., U.S. Dep’t of Homeland Sec., supra note 44, at 4.
. For an examination of the relationship between the characteristics of communities hosting detention facilities and detention outcomes, see Emily Ryo & Ian Peacock, Beyond the Walls: The Importance of Community Contexts in Immigration Detention, Am. Behav. Scientist (forthcoming 2018).
. Det. Watch Network v. U.S. Immigration & Customs Enf’t, 215 F. Supp. 3d 256, 259–60 (S.D.N.Y. 2016).
. U.S. Gov’t Accountability Office, GAO-16-231, Immigration Detention: Additional Actions Needed to Strengthen Management and Oversight of Detainee Medical Care 28 (2016), https://www.gao.gov/assets/680/675484.pdf [https://perma.cc/ZZP8-4PFW].
. Office of Inspector Gen., U.S. Dep’t of Homeland Sec., supra note 56, at 5.
. Id.; see also Am. Civil Liberties Union of N.M., supra note 109, at 35 (reporting that detainees “refused to submit grievance forms or complaints exposing officer behavior because they feared retaliation”).
. In January 2018, Ravi Ragbir was taken into ICE custody during his regular check-in with immigration authorities and flown to a detention facility in Miami without notice to his lawyer or his wife. For additional details about Ragbir’s case, see Liz Robbins, Activist Entitled to ‘Freedom to Say Goodbye,’ Judge Rules, N.Y. Times (Jan. 29, 2018), https://www.nytimes.com/2018/01/29/nyregion
. Ragbir v. Sessions, No. 18-cv-236 (KBF), 2018 U.S. Dist. LEXIS 13939, at *6–7 (S.D.N.Y. Jan. 29, 2018).
. Likewise, our measure of distance to the nearest EOIR accredited representative is time-invariant because our coding is based on the list that the EOIR published in fiscal year 2015.
. For a detailed discussion on R’s ggmap package, see David Kahle & Hadley Wickham, ggmap: Spatial Visualization with ggplot2, 5 R J. 144 (2013).
. Given the number of unique websites we reference in this Section, we do not list them here individually, but they are available upon request from the authors.