Immigration judges make consequential decisions that fundamentally affect the basic life chances of thousands of noncitizens and their family members every year. Yet, we know very little about how immigration judges make their decisions, in large part due to the scarcity of data. This lack of knowledge is especially notable when it comes to immigration bond hearings—also known as custody redetermination hearings—in which immigration judges must decide whether noncitizens should be released or continue to be detained pending the completion of their immigration cases.
Why are immigration bond decisions so important? Criminal justice studies have shown a significant relationship between pretrial detention and subsequent decisions throughout the criminal process, such as higher odds of conviction and harsher sentences (Reference DeversDevers 2011; Laura and John Arnold Foundation 2013). Studies also show that bond decisions are a critical component of criminal case processing that can have major and lasting socioeconomic consequences for criminal defendants (Open Society Foundations 2011). Immigration bond decisions are likely to be no different. For instance, a denial of bond or a prohibitively high bond amount may mean prolonged separation from families and severance from basic sources of social and economic support that might otherwise enable a noncitizen to effectively pursue legal relief from removal. And as Reference MarkowitzMarkowitz (2011: 1301–1302) has noted, for many noncitizens, removal means “life sentences of banishment” from their homes, families, and livelihoods in the United States, to countries in which they “have no family, do not speak the language, and can face serious persecution or death.”
Immigration bond hearings are also important from a broader societal perspective, because these hearings have the potential to expand or limit what some observers have described as the fastest growing—yet the least studied—type of incarceration in the United States (Reference MorehouseMorehouse 2010: 187). The number of detained noncitizens has more than doubled from approximately 200,000 in 2001, to more than 440,500 in 2013 (Reference SimanskiSimanski 2014:5). Many detainees experience lengthy periods of detention, due in part to the large volume and significant backlog of cases in immigration courts. In 2013 alone, Immigration and Customs Enforcement (ICE) is estimated to have detained over 30,000 individuals for 3 months or longer, and over 10,000 individuals for 6 months or longer (TRAC Immigration 2013). Maintaining this system of detention is costly, both economically and socially. The Department of Homeland Security (DHS) spent approximately $2 billion on immigration detention in 2014, at an average of $5.46 million per day—or $161 per detainee per day (National Immigration Forum 2014). Studies also show that immigration detention may impose enduring physical, psychological, and financial hardships on not only the detainees but also their family members, many of whom are lawful permanent residents (LPRs) or U.S. citizens (Applied Research Center 2011; Reference ChaudryChaudry et al. 2010).
Using original data on long-term immigrant detainees (defined as noncitizens detained by ICE for a continuous period of 6 months or more) held in facilities across the Central District of California, I examine for the first time judicial decision making in immigration bond hearings. I find that there are extremely wide variations in the average bond grant rates and bond amount decisions among judges in the study sample. The average grant rates ranged from 22 to 75 percent. The average bond amounts ranged from $10,667 to $80,500. What are the determinants of these decisions? My analysis shows that the odds of being granted bond are more than 3.5 times higher for detainees represented by attorneys than those who appeared pro se, net of other relevant factors. My analysis also shows that the detainees' criminal history is the only significant legally relevant factor in both the grant/deny and bond amount decisions, net of other relevant factors. This finding suggests that concerns about immigrant criminality predominate immigration bond hearings even though immigration judges, like judges in criminal proceedings, are required to consider not only whether an individual poses a “danger to the community,” but also whether he or she is a “flight risk.”
My contributions are twofold. First, the existing empirical scholarship on immigration adjudications has predominantly focused on asylum cases (see, e.g., Reference MillerMiller et al. 2015; Reference Ramji-Nogales, Schoenholtz and SchragRamji-Nogales et al. 2009). This study complicates our understanding of judicial role and decision making in immigration courts by shifting the focus of inquiry to a different type of adjudication that has grown in importance with the increasing convergence of immigration and criminal law in recent years. This convergence—what some legal scholars refer to as “crimmigration”—is a product of the growing criminalization of immigration violations on the one hand, and the expansion of criminal grounds for removal on the other (Reference StumpfStumpf 2006; see also Reference HernándezHernández 2015). Immigration bond hearings occupy a central space within this convergence in at least two ways. The ostensible goal of bond hearings is to allow immigration judges to ascertain the likelihood of the noncitizens' reappearance at later proceedings, and to ensure that they do not endanger public safety. The latter goal directly implicates one of the primary objectives of criminal punishment—incapacitation. Moreover, what is ultimately at stake in immigration bond hearings is deprivation of liberty, which is currently achieved through the same type of confinement as criminal incarceration, as discussed below. Thus, this study offers valuable insights into the increasingly important role of immigration judges in performing quasicriminal law functions of immigration law.
This study also contributes to the longstanding research on judicial decision making more generally, which has been dominated by studies of the U.S. Supreme Court and federal courts of appeals (Reference Epstein, Landes and PosnerEpstein et al. 2013: 79). I extend this body of research to deepen our understanding of how judges under conditions of extreme resource constraints and time pressure make critical decisions that potentially reinforce and reproduce social inequality. That no legally relevant factor other than the detainees' criminal history predicts immigration bond decisions suggests that heuristics might play an important role for immigration judges. The heuristics model posits that judges, like any other human beings, are “boundedly rational actors” who use mental shortcuts that “reduce the complex tasks of assessing probabilities and predicting values to simpler judgmental operations” (Reference Guthrie and GeorgeGuthrie and George 2005: 376). Whether and to what extent such decision-making dynamics might lead to errors in judgment are beyond the scope of this article. However, this study establishes an important foundation for pursuing such lines of inquiry in future studies about immigration and other courts that face significant constraints in their decisional environments.
Background
I begin with a brief description of the detention and removal processes in the United States to provide the relevant legal context for understanding immigration bond hearings. Under the Immigration and Nationality Act (INA), ICE may initiate a removal proceeding based on a noncitizen's violation of immigration laws (such as entry without inspection), or criminal convictions (such as an “aggravated felony”) that render them, including LPRs, deportable (8 U.S.C. §§ 1182, 1227). During the first stage of the removal process, the immigration judge must terminate the case if the government, represented by a DHS attorney, has not stated a valid ground for removal. If the case is not terminated, the noncitizen may seek relief from removal, such as asylum. If the judge grants relief, the noncitizen may remain in the United States; if the judge denies relief, the noncitizen will be ordered removed from the United States (8 U.S.C. § 1229a; 8 C.F.R. § 1240.8). The immigration judge's decision on the noncitizen's application for relief may be appealed to the Board of Immigration Appeals (BIA), and the BIA's decision in turn may be appealed to the federal court of appeals. The immigration courts and the BIA are part of the Executive Office for Immigration Review (EOIR), an executive agency within the Department of Justice.
Prior or subsequent to the commencement of removal proceedings, ICE has authority to detain noncitizens on a discretionary or mandatory basis. Immigration detention is not considered criminal or punitive in nature because the official purpose of immigration detention is to confine noncitizens for the “administrative purpose of holding, processing, and preparing them for removal” (GAO 2013: 8 (emphasis added)). Yet, most immigrant detention facilities “were originally built, and currently operate, as jails and prisons to confine pretrial and sentenced felons” (Reference SchriroSchriro 2009:4). For individuals held under the INA's discretionary detention provisions, ICE may release the noncitizen on conditional parole or on a bond of at least $1,500 while his or her immigration case is pending (8 U.S.C. § 1226(a)(2)).Footnote 1 The custody decision made by ICE may be appealed to the immigration court. The immigration judge's decision then may be appealed to the BIA; the BIA's custody or bond decision is final and may not be judicially reviewed (8 U.S.C. § 1226(e)).Footnote 2
Different procedures exist for noncitizens held under the INA's mandatory detention provisions. Beginning in the late 1980s, Congress enacted a series of laws closely tied to the war on drugs, mandating the detention of a certain class of noncitizens convicted of crimes, and depriving federal immigration officials of the authority to release them on bond pending their removal proceedings (Reference SayedSayed 2011: 1836–38). In 1996, the Antiterrorism and Effective Death Penalty Act (AEDPA) and the Illegal Immigration Reform and Immigrant Responsibility Act (IIRIRA) significantly broadened the use of mandatory detention by widening its net over a larger class of noncitizens (Reference JohnsonJohnson 2001). In effect, AEDPA and IIRIRA ushered in the contemporary era of mass immigration detention. Mandatory detainees include, for example, (1) certain classes of “arriving aliens,” including those seeking asylum who have not yet passed their credible fear determination and (2) noncitizens, including LPRs, convicted of certain crimes enumerated in the INA (8 U.S.C. §§ 1225(b), 1226(c)). Noncitizens in removal proceedings with triggering criminal offenses are typically detained by ICE after they have already served their jail/prison terms, if any. Recently, however, a class action lawsuit, Rodriguez v. Robbins, brought by long-term detainees in the Central District of California, changed this legal landscape by requiring the government to provide bond hearings to noncitizens, including mandatory detainees, who have been continuously detained for 180 days or more.
As the data for this study comes from Rodriguez class members, I briefly discuss the history and holding of Rodriguez. In 2007, Alejandro Rodriguez, who had been in ICE custody for more than 3 years, and similarly situated noncitizens in the Central District of California, filed a class action lawsuit challenging the legality of detention lasting more than 6 months without individualized bond hearings before an immigration judge. The Rodriguez class formally consists of all noncitizens in ICE custody within the Central District of California who: (1) are or were detained for longer than 6 months pursuant to one of the general immigration detention statutesFootnote 3 pending completion of removal proceedings, including judicial review, (2) are not and have not been detained pursuant to a national security detention statute, and (3) have not been afforded a hearing to determine whether their detention is justified (Rodriguez v. Robbins 2013:Note 1).
The District Court in Rodriguez granted a preliminary injunction, holding that the immigration judge must release these detainees “on reasonable conditions of supervision, including electronic monitoring if necessary, unless the government shows by clear and convincing evidence that continued detention is justified based on his or her danger to the community or risk of flight” (Rodriguez v. Robbins 2012). The Ninth Circuit Court of Appeals (Rodriguez v. Robbins 2013) affirmed the District Court's decision, and the District Court subsequently issued a permanent injunction requiring bond hearings for all Rodriguez class members (Rodriguez v. Robbins 2013).Footnote 4 With this legal background in mind, I now develop a theoretical framework for analyzing immigration bond decisions.
Theoretical Framework
What are the determinants of immigration judges' bond decisions? Reference GilboyGilboy's (1987) examination of one immigration court in Chicago is the only empirical study of immigration bond hearings to date. Gilboy's study, however, does not contain case-level data that allows for an analysis of the determinants of immigration bond decisions. Moreover, Gilboy's study predates the 1996 amendments to the INA that dramatically changed the legal landscape and set in motion the contemporary practices of immigration detention in the United States. Thus, to theorize about the determinants of immigration bond decisions, I turn to research on immigration courts generally, and research on criminal pretrial custody hearings.
Research on Immigration Courts
I begin by highlighting certain basic characteristics of immigration courts and judges (see Reference BaumBaum 2010 for a helpful review) to contextualize my review of the relevant findings from research on immigration courts. First, immigration judges do not derive their authority from Article III of the U.S. Constitution, which establishes the Judicial Branch. Instead, an immigration judge is an attorney whom the U.S. Attorney General appoints as an administrative judge pursuant to the civil service laws (8 U.S.C. § 1101(b)(4)). Thus, immigration judges do not enjoy life tenure and can be removed from the bench for misconduct or reassigned to another position at the discretion of the Attorney General (Reference LegomskyLegomsky 2006: 373–74). Nonetheless, immigration judges are formally expected to exercise a high degree of “independent judgment and discretion” under the INA (8 C.F.R. § 1003.10). Second, immigration judges face an extremely large volume and significant backlog of cases. In 2010, federal district court judges had an average pending caseload of about 400 cases, with each judge typically maintaining three law clerks to assist the judge; in comparison, immigration judges had an average pending caseload of about 1,500 cases, with one law clerk shared among four judges (Reference MarksMarks 2012: 27).Footnote 5 In 2014, more than 408,000 cases were pending in immigration courts across the United States, which is 240,000 more than the total number of pending cases in 2004 (TRAC Immigration 2015).
Empirical studies of immigration adjudications are relatively new and predominantly focused on asylum cases (see, e.g., GAO 2008; Reference Keith, Holmes and MillerKeith et al. 2013; Reference Ramji-Nogales, Schoenholtz and SchragRamji-Nogales et al. 2009; Reference Rottman, Fariss and PoeRottman et al. 2009). I highlight two major findings from this line of research. First, consistent with the attitudinal model of judicial decision making, which posits that judging is a matter of personal beliefs and politics (Reference Segal and SpaethSegal and Spaeth 1993; Reference SunsteinSunstein 2006), a number of studies have found that disparities in asylum outcomes are significantly associated with immigration judges' personal characteristics. For example, Reference Ramji-Nogales, Schoenholtz and SchragRamji-Nogales et al. (2009), in their groundbreaking study of disparities in asylum grant rates, show that the chances of winning asylum are significantly associated with the gender of the immigration judge, and the immigration judge's work experience prior to appointment to the bench. Similarly, Reference Keith, Holmes and MillerKeith et al. (2013) find that policy predispositions of immigration judges, as measured by an index of past career experiences, interact with legal and extralegal factors to explain wide variations in asylum grant rates.
The second notable finding from the emerging research on immigration adjudications relates to the importance of legal representation. Unlike criminal defendants, noncitizens do not have a right to government-appointed counsel (8 U.S.C. § 1229a(b)(4)(A)). As a result, most noncitizens in immigration proceedings lack legal representation (New York Immigration Representation Study 2011: 358). For example, in a recent national study of over 1.2 million immigration removal cases decided between 2007 and 2013, Reference Eagly and ShaferEagly and Shafer (2015) find that only 37 percent of noncitizens had legal representation. This gap in legal representation in immigration proceedings has caused growing public concern, as studies have shown a significant association between legal representation and favorable case outcomes. For example, Eagly and Shafter find that represented noncitizens are more likely to have their cases terminated, more likely to seek relief from removal, and more likely to obtain the relief they seek. In another recent study focusing on asylum cases, Reference MillerMiller et al. (2015) show that whether or not the noncitizen is granted asylum is significantly related to the quality of legal representation.
Taken together, this body of research suggests that personal characteristics of immigration judges and legal representation might play important roles in shaping immigration bond decisions. Even so, bond decisions differ from asylum and removal decisions—both in terms of the type of legal and factual questions raised, and the average duration of the hearings. Thus, I now turn to research on criminal pretrial custody hearings, which are likely to be closer analogs to immigration bond hearings than asylum or removal proceedings.
Research on Criminal Pretrial Custody Decisions
The research on criminal pretrial custody decisions is varied and longstanding (for earlier studies, see, e.g., Reference Bock and FrazierBock and Frazier 1977; Reference Goldkamp and GottfredsonGoldkamp and Gottfredson 1979; Reference SuffetSuffet 1966). My discussion of this body of research focuses on quantitative studies examining pretrial custody decisions in U.S. courts. As Reference CrossCross (2007: 47) has argued, “The law is the obvious possible explanation for [judicial] decisions.” Thus, the natural starting point for analysis in many studies of criminal pretrial custody hearings has been what scholars call the legal model of judicial behavior. There are variants of the legal model, but one core idea they share is that judges base their decisions on the standards set forth in the law (e.g., the constitution, statutes, legal precedent, and court rules) (Reference Segal and GeyhSegal 2011: 18–19). Accordingly, many studies of criminal pretrial custody decisions have assessed the significance of various legal factors—statutorily prescribed guidelines—that judges are supposed to consider in their decision making. Studies have also focused on the effects of extralegal factors on hearing outcomes. In using the term “extralegal” in this study, I follow Reference Nagel3:482), who defined the term in her study of pretrial release as those factors that are “not specifically prescribed in the relevant … law”; Nagel explicitly eschewed the definition of extralegal factors as those that are “illegal,” “inappropriate,” or “socially unjust.”
On the whole, empirical studies on legal factors associated with pretrial decisions generally find that the seriousness of current charges and prior criminal records are the most consistent predictors of pretrial decisions, though findings vary depending on the types of decisions examined (e.g., the grant/deny decision versus the bond amount decision) (see, e.g., Reference Bock and FrazierBock and Frazier 1977; Reference Goldkamp and GottfredsonGoldkamp and Gottfredson 1979; Reference Gottfredson and GottfredsonGottfredson and Gottfredson 1990; Reference NagelNagel 1983; Reference SpohnSpohn 2009). The study findings are more mixed with respect to the effects of other legal factors on pretrial custody decisions. While some studies find that community ties, employment and financial conditions, and family ties, have little to no significant relationship to pretrial custody decisions, other studies offer contrary evidence (compare Reference DalyDaly 1987: 164 with Reference Goldkamp and GottfredsonGoldkamp and Gottfredson 1979: 240). Thus, findings on the effects of legal factors on criminal pretrial custody decisions are relatively mixed and difficult to generalize.
Likewise, study findings on the effects of extralegal factors on pretrial custody decisions are also mixed. The effect of offenders' race on pretrial custody decisions has been the focus of growing empirical inquiry (see, e.g., Reference AlbonettiAlbonetti et al. 1989; Reference Ayres and WaldfogelAyres and Waldfogel 1994; Reference FreiburgerFreiburger et al. 2010), but the results are varied.Footnote 6 In one study, Reference FreeFree (2002: 206–10) concluded that racial disparities in bond decisions are significant and have been stable over time, especially with respect to bond amounts. But in another study, Reference FreeFree (2004) concluded that black offenders living in cities with less than 10 percent black population typically do not experience racial disparities in bond decisions, presumably because whites do not perceive a small number of blacks as a threat. Seemingly conflicting empirical findings in this area of research may be due in part to varying legal rules and norms in different areas of the country. For example, in a study of bond decisions in New Haven courts, Reference Ayres and WaldfogelAyres and Waldfogel (1994) found that black defendants are given higher bond amounts than similarly situated white defendants. In contrast, Reference Katz and SpohnKatz and Spohn (1995) examined defendants charged with violent felonies in Detroit and found no racial disparities in bond amounts. Most recently, in a national study of felony defendants, Reference McIntyre and BaradaranMcIntyre and Barandaran (2013: 769) argued that racial gaps in bond decisions are “entirely accounted for, on average, by differing probabilities of rearrest for violent crime.”
I conclude this review of the research on criminal pretrial custody hearings by turning to studies on the heuristics model of decision making. Although there has been a growing interest in heuristics to explain judicial behavior (see, e.g., Reference Bainbridge and GulatiBainbridge and Gulati 2002; Reference Guthrie, Rachlinski and WistrichGuthrie et al. 2001; Reference Rachlinski, Wistrich and GuthrieRachlinski et al. 2013), heuristics have received scant attention in research on pretrial custody decisions, with the exception of two studies I discuss below. Heuristics are mental shortcuts or rules of thumb in reasoning that people commonly use to make decisions (Reference Shah and OppenheimerShah and Oppenheimer 2008). The heuristics research is built on assumptions of “bounded rationality,” which rejects the conventional rational-choice fiction of unbounded time, knowledge, and computational power presumed to be available to individuals (Reference KahnemanKahneman 2003).
One school of bounded rationality focuses on “fast and frugal” heuristics (see, e.g., Reference Gigerenzer, Koehler and HarveyGigerenzer 2004).Footnote 7 Fast and frugal heuristics eschew optimization, which requires weighing and integrating all available cues, in favor of simple informational search and stopping rules, and one-reason decision making (Reference Gigerenzer and GoldsteinGigerenzer and Goldstein 1996). Consistent with the fast and frugal model, Reference Dhami and AytonDhami and Ayton (2001) found in their study of bond decisions in England and Wales that the magistrates' decisions on hypothetical cases were based on a single cue (though individual magistrates differed as to which cue they used). In another study of over 340 bond decisions made by judges in London, Reference DhamiDhami (2003) again found that the judges relied on a single cue—the previous decisions made by the police, previous bench, or the prosecutor (e.g., whether prosecution requested conditional bond or opposed bond). Dhami thus concluded that simple heuristics in which judges appear to be “passing the buck” were the best predictors of criminal bond decisions in her study.
My analysis of the determinants of immigration bond decisions builds on and extends the research on immigration courts and the research on criminal pretrial custody hearings. Together, these two bodies of research suggest that a useful starting point for analyzing immigration bond decisions is to examine the effects of legal and extralegal factors (including whether or not the detainee had an attorney). In addition, research on immigration courts suggests that judge characteristics might be significantly related to bond decisions. Finally, research on heuristics of judging casts some doubt on the applicability of the legal model for immigration bond decisions, given the substantial cognitive and informational constraints, and time pressures that immigration judges face.
Data and Method
Data
The data for this study comes from original survey data collected from class members of Rodriguez. Between May 2013 and March 2014, in-person surveys were conducted with 565 detainees at four detention facilities in the Central District of California. These facilities are the James A. Musick Facility (Musick), the Theo Lacy Facility (Theo Lacy), the Santa Ana City Jail (Santa Ana), and the Adelanto Detention Facility (Adelanto). Musick and Theo Lacy are county jails operated by the Orange County Sheriff's Department. Santa Ana is a city jail operated by the Santa Ana Police Department. ICE contracts with Musick, Theo Lacy, and Santa Ana to hold immigrant detainees.Footnote 8 Adelanto is operated by a private prison company, the GEO Group, and houses only immigrant detainees. Approximately 23 percent of the survey respondents were held at Musick; 21 percent at Theo Lacy; 13 percent at Santa Ana; and 43 percent at Adelanto. All survey respondents were 18 years of age or older, as juveniles are not covered by the injunction in Rodriguez.Footnote 9
The surveys were orally administered in person in English or Spanish, depending on the respondents' preference. The surveys lasted about 60 minutes on average. The respondents were not paid for participating in the survey. In addition to the bond hearing information, the survey captures information about the detainees' background, including their demographic profile, immigration and criminal history, predetention employment history, and household and family relationships. Class members of Rodriguez were provided detailed information about the survey and only those who consented to participate were surveyed. More than 92 percent of the detainees who were provided information about the survey by the interviewers completed the survey. There were no significant differences in refusal rates by gender or country of origin. If a respondent had not yet had their bond hearing at the time of the survey or if the bond hearing had been continued for any reason, the interviewers attempted to survey the respondent again at a later date to capture the bond hearing information. In total, 113 respondents were reinterviewed for this purpose. Given that the survey took place in detention facilities, it is possible that the survey sample may overrepresent class members who were denied bond, or were granted bond but could not post it at the time of the survey.
A series of structured courtroom observations provided useful insights that guided the wording and the ordering of survey items—particularly those relating to the bond hearings. The courtroom observational data contains information about 40 substantive bond hearings held in five courtrooms in the Central District of California—three courtrooms at Adelanto and two in downtown Los Angeles. Although the observed hearings pertained to different Rodriguez class members than those who participated in the survey, it is unlikely that the observed hearings varied systematically from the hearings captured in the survey. The research team used a detailed coding instrument to collect basic information about each hearing, and major portions of each hearing were transcribed verbatim as much as possible during and immediately after the hearings. These courtroom observations also provided me with basic insights about bond hearing dynamics that informed my interpretation of the analysis results presented below. For example, our observations of attorneys in action in the courtroom were important in my consideration of the possible signaling functions of legal representation.
To understand how bond hearings are assigned to immigration judges, I filed a Freedom of Information Act request with the EOIR. The EOIR response indicated that what is “normally done” is “equal distribution” of cases across judges responsible for the detained docket. The EOIR response, however, also suggested that when the immigration court became overwhelmed with the number of bond hearings, certain judges were reassigned from the nondetained to the detained docket to preside solely over bond hearings. There is nothing to indicate in the EOIR response that case assignments were dependent on individual case or detainee characteristics. I also corresponded with a number of local immigration attorneys in law school clinics, nonprofit legal services, and private practice who regularly represent detainees in immigration courts, and their general consensus was largely consistent with this conclusion.
It is difficult to determine how representative Rodriguez class members are of the entire immigrant detainee population in the United States, since the DHS does not make publicly available detailed demographic and case-related information about immigrant detainees in the United States. However, there are two general characteristics of Rodriguez class members that may be worth noting. First, the Rodriguez class members on average are more likely to be contesting their removability and/or seeking relief from removal compared to detainees who experience short-term detention (detention lasting less than six months).Footnote 10 Second, Rodriguez class members on average are more likely to have a criminal record than short-term immigrant detainees, as many of the former are mandatorily detained and placed in removal proceedings due to their triggering criminal offenses. As shown in Supporting Information Table A1, the two most common convictions among the respondents in the effective sample are, by far, traffic and drug-related (45 percent, respectively).
My unit of analysis is bond hearings; only one hearing per respondent is included in the analysis. I analyze only the first substantive bond hearing held around or after 6 months of detention. “Substantive hearings” are hearings that resulted in either a grant or a denial of bond. Thus, I exclude all hearings that were continued, and hearings in which the immigration judge held there was no jurisdiction to hold the hearing. These exclusions reduced the sample from 565 to 526.Footnote 11 Approximately 48 percent of the 526 respondents were seeking asylum, withholding of removal, or relief under the United Nations Convention against Torture; 49 percent were seeking cancellation of removal or adjustment of status; 10 percent were seeking relief under Violence against Women Act, U Visa, or T visa; 4 percent were seeking voluntary departure; and 6 percent were seeking some other form of legal relief.Footnote 12 After deleting cases that were missing on one or more of the variables included in the regression models,Footnote 13 the effective sample for the bond grant/deny analysis became 449, and 261 for the bond amount analysis.Footnote 14
Measures
Supporting Information Table A3 shows all of the variables used in the analyses and their corresponding survey items. I discuss each of these variables in turn below.
Dependent Variable
I analyzed two major bond hearing outcomes. The first outcome variable is the bond grant/deny decision (1=granted; 0=denied). The second outcome variable is the bond amount for those who were granted bond.
Independent Variables
There are two major categories of independent variables. The first category includes extralegal factors—the basic detainee background characteristics. The second category includes legally relevant factors.
Detainee Background Characteristics.
The detainee background characteristics include gender, age, English language ability, race, educational level, and whether or not the detainee had an attorney at the bond hearing.
Legally Relevant Factors.
As in criminal bond hearings, judges in immigration bond hearings are required to determine whether the individual poses a danger to the community or is a flight risk. The following factors are enumerated in the Immigration Judge BenchbookFootnote 15 as relevant to this determination: (1) whether the noncitizen has a fixed address in the United States; (2) the length of residence in the United States; (3) family ties in the United States, particularly those who can confer immigration benefits on the noncitizen; (4) employment history in the United States, including length and stability; (5) immigration record; (6) prior attempts to escape authorities or other flights to avoid prosecution; (7) prior failures to appear for scheduled court proceedings; and (8) criminal record, including extensiveness and recency, indicating consistent disrespect for the law and ineligibility for relief from deportation/removal (EOIR 2015). In general, the immigration judge has “broad discretion in deciding the factors that he or she may consider in custody redeterminations,” and the judge may give greater weight to certain factors “as long as the decision is reasonable” (In re Guerra 2006:40).
These legally relevant factors are captured by the following items in the survey: whether or not the respondent had lived at his or her own home or family member's home pre-detention; the number of years of stay in the United States; whether or not the respondent has a U.S. citizen or LPR child or spouse; whether or not the respondent had been employed within the six month period prior to being detained; current legal status; whether or not the respondent's immigration case is before the Ninth Circuit Court of Appeals; whether or not the respondent had been previously deported from the United States; history of obstruction of justice; and criminal history. To ensure the accuracy of the criminal history data, the following protocol was implemented: (1) the criminal history items were placed well into the survey to enable the interviewers to first develop a strong rapport with the respondents before administering those items, (2) before asking about criminal history, the interviewers reminded the respondents that all information obtained would remain anonymous and confidential, (3) the interviewers assured the respondents that the survey was not seeking details about the situations leading up to their arrests/convictions; rather, the goal was to find out what the government thinks the respondent did, and (4) the interviewers first asked the respondents about the total number of past arrests/convictions, and then used a grid to capture more detailed information about each arrest/conviction.Footnote 16
Supporting Information Table A4 lists each of the legally relevant factors and their corresponding variables included in the analyses. As explained earlier, because all respondents in the effective sample have at least one type of criminal conviction, the “0” category for each criminal-conviction dummy variable represents respondents with criminal conviction(s) other than the conviction coded as “1.” For example, the “0” category for Convicted of Sex-Related Offense consists of all respondents who have convictions other than sex-related convictions.
Analytical Strategy
I performed two main sets of regression analyses using the variables defined in Supporting Information Table A3. First, I examined the determinants of bond grant/deny decisions. Since the outcome variable in this analysis is binary (1=granted; 0=denied), I employed logistic regression to estimate the effects of various extralegal and legally relevant factors on decisions to grant or deny bond. Second, I examined the determinants of bond amount decisions for those detainees who were granted bond. Bond amount is truncated at the minimum bond amount mandated by law ($1,500), and highly skewed with variance exceeding the mean. Thus, I employed truncated negative binomial regression to estimate the effects of various extralegal and legally relevant factors on bond amounts. Negative binomial regression is a generalization of Poisson regression with an extra parameter to model overdispersion. Finally, I also performed a separate series of bivariate tests to examine the relationships between basic judge characteristics, such as gender, and the two types of bond decisions, respectively.
In this study, I conceptualize bond grant/deny and bond amount outcomes as two analytically distinct decisions. This approach is consistent with the widespread and longstanding tradition in studies of criminal pretrial custody hearings that I reviewed above. However, it is possible that in practice, some immigration judges may be setting bond amounts at levels that constitute effective denials of bond. If judges were systematically engaged in effective denials, we might expect a significant positive relationship between bond grant rates and bond amounts (that is, judges are granting bond at a high rate, only to set very high bond amounts). My analysis, however, showed a negative correlation of −0.45 (p < 0.001) between bond grant rates and average bond amounts.
It is possible that the attorney variable is endogenous. That is, detainees with legal representation may be systematically different from those who lack representation, and those differences might be related to hearing outcomes. For example, it is possible that attorneys might be strategically selecting cases that are more likely to win (but see Reference MillerMiller et al. 2015: 220 (finding no evidence of such selection effects in asylum cases)). To examine this possible issue, I explored recursive bivariate probit regressions (see Reference MaddalaMaddala 1983; Reference Marra and RadiceMarra and Radice 2011) with two simultaneous equations: (1) outcome equation predicting the bond grant/deny decision; and (2) selection equation predicting whether or not the detainee had legal representation. Facility location, included in the selection equation, may be a suitable instrument on the grounds that it does not have a direct effect on the grant/deny decision, but affects the likelihood of legal representation since attorneys are less likely to take cases that require travel to remote locations. However, I encountered computational issues in performing the bivariate probit regression analysis, likely due to the sample size (see Reference Monfardini and RadiceMonfardini and Radice 2008: 272). Thus, I was unable to determine with the available data whether legal representation is endogenous.
Results
Descriptive and Bivariate Patterns
I begin by examining the bond grant rates. The base sample contains 526 bond hearings presided by 20 different judges. Figure 1 shows the grant rates of individual judges in the study sample who presided over 15 or more cases. The “other” category is a residual category that includes 11 judges who presided over less than 15 cases, and judges whom the respondent could not identify by name. As shown in Figure 1, there are significant differences in the grant rates across judges, ranging from 22 to 75 percent. Figures 2 and 3, respectively, show mean and median bond amounts for each judge. According to Figure 2, the mean bond amount ranges from $10,667 to $80,500. The median bond amount range is narrower than the mean bond amount range, but still is relatively large at $10,000 to $32,500, as shown in Figure 3.
Figure 1. Bond Grant Rate by Judge
Figure 2. Mean Bond Amount by Judge
Figure 3. Median Bond Amount by Judge
Next, to begin to explore what factors might explain these decisions, I examine bivariate test results. Table 1 presents the means, the standard deviations, and the ranges for all variables used in the analysis of grant/deny decisions. Univariate statistics show that the respondents in the effective sample are overwhelmingly Latino men, nearly half of whom have a high school degree or more. About 70 percent of the sample is currently in the United States without authorization and the remaining 30 percent are LPRs or have some other valid visa. The sample includes individuals with long and established ties to the United States and stable housing and employment histories. The survey respondents have been in the United States for an average of 20 years, and almost 70 percent have a U.S. citizen or LPR spouse or child living in the United States. In the 6 months prior to ICE detention or jail/prison time leading up to their ICE detention, 84 percent were living in their own homes or with family members, and 90 percent reported having a job.
Table 1. Descriptive and Bivariate Statistics for Variables Used in the Analysis of Grant/Deny Decisions
Source: Rodriguez Survey, 2013–2014; N=449.
Notes: a For binary variables, bivariate tests consist of chi-square tests; for continuous variables, bivariate tests consist of t tests. bReference category is “Speaks English Just a Little/Not at All.” cCriminal offenses were categorized following the classification system used in GAO 2011, Table 7, which is based on the FBI classification scheme. All categories include attempt and conspiracy to commit the respective offense. For specific offenses included in each conviction variable, see Supporting Information Table A3.
* p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed tests).
The final column of Table 1 shows bivariate test results comparing bond grant rates for each of the independent variables. Extralegal factors are generally not significantly related to grant/deny decisions. One exception is that detainees with an attorney at their bond hearings are much more likely to be granted bond. The only legally relevant factors significantly related to bond grant/deny decisions are those pertaining to the detainees' criminal history. Detainees with more recent convictions, and drug and property convictions (compared to other types of convictions) are more likely to be granted bond. On the other hand, detainees with felony convictions, DUI convictions, and sex-related convictions (compared to other types of convictions) are much less likely to be granted bond.
Table 2 shows bivariate statistics for the variables used in the analysis of bond amounts. As with bond grant/deny decisions, there are few statistically significant differences in bond amounts by extralegal factors, with the exception of educational levels. Those with a high school degree or more are given higher bond amounts on average, compared to those who did not receive a high school degree. Table 2 also shows that legally relevant factors generally are not significantly related to bond amounts; one exception is the number of felony convictions. A separate analysis of a dichotomous felony variable (not shown in Table 2) indicates that detainees with felonies have an average bond amount of $47,133 while those without felonies have an average bond amount of $20,040 (p < 0.001).
Table 2. Bivariate Statistics for Variables Used in the Analysis of Bond Amount Decisions
Source: Rodriguez Survey, 2013–2014; N=261.
Notes: a Amounts under X=1 are mean amounts for binary variables at values of 1 (e.g., male, Hispanic or Latino/a). Amounts under X=0 are mean amounts for binary variables at values of 0 (e.g., female, non Hispanic or non Latino/a). bFor binary variables, bivariate tests consist of t tests; for continuous variables, bivariate tests consist of Spearman correlation tests. cReference category is “Speaks English Just a Little/Not at All.” dCriminal offenses were categorized following the classification system used in GAO 2011, Table 7, which is based on the FBI classification scheme. All categories include attempt and conspiracy to commit the respective offense. For specific offenses included in each conviction variable, see Supporting Information Table A3.
* p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed tests).
Finally, I conducted a series of bivariate tests between number of basic judge characteristics and grant/deny and bond amount decisions, respectively. The judge characteristics that I examined were gender, the political party of the appointing U.S. Attorney General, and the type of prior work experience (e.g., having worked for the government, the DHS, and/or nongovernmental organizations, and/or having been in private legal practice). The results (not shown but available on request) of these bivariate tests indicate that none of the judge characteristics are significantly related to the bond decisions. These results, however, must be interpreted with caution given that the current number of judges in the data may not allow for sufficient statistical power to detect significant effects. In addition, the key factor in the attitudinal model is the political ideology of judges; here, I have only a crude proxy for the immigration judges' political ideology—the political ideology of the appointing Attorney General. Given the lack of significant results on bivariate tests, I did not perform multivariate analyses using these variables.
Multivariate Analyses
Hearings are nested within judges; however, there is an insufficient number of clusters and observations within clusters to reliably estimate mixed level models (see Reference Moineddin, Matheson and GlazierMoineddin et al. 2007) or to adjust the standard errors for clustering (see Reference Cameron, Gelbach and MillerCameron et al. 2008). Thus, I performed judge fixed effects regressions by including effect-coded judge variables in the full models (see Model 3 in Tables 3 and 4, respectively). Effect coding in this situation allows for a more meaningful interpretation than dummy coding, given that each of the effect-coded judge variable's coefficient reflects the difference between the mean of the judge coded 1 and the grand mean. Of note, I examined intraclass correlation (“ICC”) for each dependent variable. If all variation in the sample across the outcome variable is due to clustering of hearings within judges, the ICC would equal one. Conversely, if none of the variation in the sample is due to clustering of hearings within judges, the ICC would equal zero. For grant/deny decisions, the ICC was 0.09, indicating that about nine percent of the variation in the sample is attributable to clustering of hearing within judges. For the bond amount decisions, the ICC was 0.06, indicating that about six percent of the variation in the sample was attributable to clustering of hearings within judges. Neither of the ICCs was significantly greater than zero at p < 0.05.
Table 3. Odds Ratios from Logistic Regression Models Predicting Bond Grant/Deny Decisions
Source: Rodriguez Survey, 2013–2014; N=449.
Notes: Standard errors in parentheses. aReference category is “Speaks English Just a Little/Not at All.” bReference category is “Lawful Permanent Resident.” cCriminal offenses were categorized following the classification system used in GAO 2011, Table 7, which is based on the FBI classification scheme. All categories include attempt and conspiracy to commit the respective offense. For specific offenses included in each conviction variable, see Supporting Information Table A3.
* p < 0.05; ** p < 0.01; *** p < 0.001 (two-tailed tests).
Table 4. Incidence Rate Ratios from Negative Binomial Regression Models Predicting Bond Amount Decisions
Source: Rodriguez Survey, 2013–2014; N=261.
Notes: Standard errors in parentheses. aReference category is “Speaks English Just a Little/Not at All.” bReference category is “Lawful Permanent Resident.” cCriminal offenses were categorized following the classification system used in GAO 2011, Table 7, which is based on the FBI classification scheme. All categories include attempt and conspiracy to commit the respective offense. For specific offenses included in each conviction variable, see Supporting Information Table A3.
* p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed tests).
The three models in Tables 3 and 4, respectively, are nested. Model 1 contains only the detainee background characteristics. Model 2 includes legally relevant factors. Model 3, the full model, contains effect-coded judge variables. Table 3 shows the results of a set of binomial logistic regression models, with bond grant/deny decisions as the outcome variable. For ease of interpretation, I present the results in odds ratios rather than coefficient estimates. The odds ratios represent the odds of being granted bond versus being denied bond. An odds ratio higher than 1 indicates an increase in the odds associated with a one-unit increase in a given independent variable. An odds ratio between 0 and 1 indicates a decrease in the odds associated with a one-unit increase in a given independent variable. According to Wald tests, Model 3, the full model, fits the data significantly better than Model 2 or Model 1. The following discussion focuses on Model 3.
Table 3 shows two major patterns of results. First, extralegal factors are not significant across any of the models, with one exception. Model 3 shows that the odds of being granted bond are more than 3.5 times higher for detainees with attorneys than those who appeared pro se. The significant attorney effect, however, should be interpreted with caution, since the attorney variable may be endogenous, as discussed earlier. Second, there is a clear pattern of results with respect to the legally relevant factors. The only significant legally relevant variables are those relating to the detainees' criminal history. Model 3 shows that for each additional felony conviction, the odds of being granted bond are reduced by more than half ([1 − 0.435] × 100). DUI convictions also reduce the odds of being granted bond by about half. Sex-related convictions reduce the odds of being granted by over two-thirds (relative to non sex-related convictions), while property convictions nearly double the odds of being granted bond (relative to non property convictions). Finally, Model 3 shows that the odds of granting bond are significantly different across a number of judges, even after controlling for detainee background characteristics and legally relevant factors.
I turn next to Table 4, which analyzes bond amount decisions. The dispersion parameter (alpha in Stata) is significantly greater than zero (p < 0.001, not shown in Table 4), indicating that the negative binomial model is more appropriate than the Poisson model. All results in Table 4 are shown as incidence-rate ratios (IRR), which are defined as exp(β), where β is a log count. An IRR of greater than 1 indicates an increase in the likelihood of getting a higher bond amount for each one-unit increase in a given independent variable. An IRR between 0 and 1 indicates a decrease in the likelihood of getting a higher bond amount for each one-unit increase in a given independent variable. According to Wald tests, Model 3, the full model, fits the data significantly better than Model 2 or Model 1.
Overall, Model 3 of Table 4 shows that after the inclusion of effect-coded judge variables, the only variable that remains significant is the number of felony convictions. For each additional felony conviction, the IRR increases by a factor of 0.37. A more intuitive way of interpreting this result is to examine the predicted bond amounts with incremental increases in the number of felonies. I calculated these predicted amounts using the margins command in Stata. With all other variables held at their observed values, Model 3 predicts that on average, respondents with no felonies will receive a bond amount of about $21,753, whereas those with one felony will receive a bond amount of $29,807 and those with two felonies, a bond amount of $40,841. Of note, although the attorney variable is significant in Models 1 and 2 of Table 4, this variable loses its significance when effect-coded judge variables are included in Model 3. Finally, Model 3 shows that differences between individual judges' average bond amounts and the grand mean are statistically significant for all but two judges, net of detainee background characteristics and legally relevant factors.
Discussion and Conclusion
Despite the critical importance of immigration bond hearings for both the detained population and the broader public, no previous study has systematically examined immigration bond hearings. This study fills this gap using original survey data to analyze the effects of extralegal and legally relevant factors on the bond grant/deny and bond amount decisions. My descriptive analysis shows that there are extremely wide variations in bond grant rates and bond amount decisions among judges in the study sample. The multivariate analyses examining the determinants of these decisions produced two major findings. First, consistent with the existing research on legal representation and immigration case outcomes, detainees who had attorneys were significantly more likely to be granted bond compared to those who lacked legal representation. Second, I find that there is only one significant determinant of both the grant/deny and bond amount decisions—the detainees' criminal history.
I conclude by discussing the major implications of these findings and key directions for future research. Is there a causal relationship, direct or indirect, between legal representation and the immigration judges' bond decisions? More generally, how, under what conditions, and in what stages of removal proceedings, do immigration attorneys impact their clients' legal outcomes? Can nonattorney advocates confer the same level and type of advantages as attorneys in immigration bond hearings, and immigration proceedings more generally? This study's multivariate analysis, which shows a substantial and significant relationship between legal representation and the likelihood of being granted bond, raises these and other related questions that are critical to resolving broader debates about access to justice in immigration courts. Below, I suggest some promising lines of inquiry for future research in this area by proposing a number of ways in which legal representation might matter in bond hearings.
According to the U.S. Supreme Court, “immigration laws have been termed second only to the Internal Revenue Code in complexity,” and “a lawyer is often the only person who could thread the labyrinth” (Castro-O'Ryan v. INS 1988:1312). We might thus expect an attorney's legal expertise, including his or her judge-specific knowledge (see Reference MillerMiller et al. 2015), to have a direct impact on bond decisions. However, attorneys might also shape bond decisions in less direct, albeit equally important, ways. Immigration detention facilities are “total institutions” (Reference GoffmanGoffman 1961) that are segregated from the outside world by walls and fences, as well as rules that restrict outside contact. In such an environment, attorneys may serve as a critical intermediary between the detainee and the outside world, allowing the detainee to obtain relevant documents and evidence in support of his or her case.
Moreover, immigration attorneys, by virtue of their professional and “repeat player” status (Reference GalanterGalanter 1974), can negotiate with government attorneys in advance of the bond hearing to propose a “bond settlement” to immigration judges. These proposals may serve as the default starting point for a judge's decision-making process, or they may even allow the judge to bypass his or her own decision making. In the courtroom observational data, a number of such settlement negotiations occurred between the attorneys off the record shortly before the start of the bond hearings. In those cases, the immigration judges did not pose substantive questions to the detainees nor to the attorneys; rather, the judges simply adopted the proposed bond amount agreed upon by the attorneys. By contrast, no government attorney ever approached a detainee directly to negotiate a bond settlement. In short, detainees who lack legal representation might be categorically excluded from this negotiation process that could increase the likelihood of being granted bond.
In addition, immigration attorneys might perform an important signaling function during the bond hearings. As Garvin, a detainee whom I interviewed after his bond release explained, an immigration judge who sees a detainee with an attorney at the bond hearing might view that detainee as a “worthy opponent”—one who might be deemed more deserving of attention, respect, and opportunity by immigration judges. In addition, attorneys might also serve as a signal to the immigration judges that the detainee is relatively more committed to and invested in the legal process, making him or her appear less of a flight risk than pro se detainees. In Garvin's words: “A lot of [the bond outcome] has to do with the fact that if you acquire an attorney, it shows that you are serious about your case.”Footnote 17
This study's other major finding, that the detainees' criminal history is the only significant legally relevant factor in bond decisions, suggests that the “mark of a criminal record” (Reference PagerPager 2003) casts a long and decisive shadow in immigration bond hearings. The current data does not allow me to analyze the accuracy of the immigration judges' bond decisions. However, evidence is mixed in the criminal justice context on whether criminal history is a reliable predictor of pretrial misconduct, including failures to appear and rearrests. As Reference Baradaran and McIntyreBarandaran and McIntyre (2012:500) explain, there is disagreement in existing research “as to whether the current charge or past convictions are relevant as predictors of future crimes, whether flight risk is linked to pretrial violence, and whether judges are accurately able to predict which defendants are dangerous.” This uncertainty in our knowledge about the predictive power of criminal history in the criminal justice context raises an important policy question about whether and to what extent immigration judges should have the discretion to rely so heavily on any one legally relevant factor in making bond decisions.
Why might the criminal history of detainees be the only significant legally relevant factor in immigration bond decisions? I highlight a number of possible explanations that warrant more systematic theoretical and empirical investigation in future research. I begin by highlighting a number of immediate contextual factors that might lead immigration judges to focus on the criminal history of detainees. I then consider the broader contemporary legal and policy environment in which immigration courts operate—an environment that increasingly emphasizes the crime control imperatives of immigration law.
One defining characteristic of the immigration judges' job is the surging caseloads and chronic lack of resources in immigration courts. For example, immigration judges have reported a shortage of law clerks and language interpreters, and failing computers and equipment for recording the hearings (Reference LustigLustig et al. 2008; Reference MarksMarks 2012). Immigration judges have also pointed to the intense pressures they face to move cases off their dockets (Reference Benson and WheelerBenson and Wheeler 2012:24). One immigration judge, describing this pressure, noted in a recent survey: “There is not enough time to think” (Reference LustigLustig et al. 2008:66). Yet, immigration judges must render extemporaneous oral decisions requiring difficult probabilistic calculations. Under these conditions of time and resource constraints, immigration judges may construe criminal history as the most efficient and determinate proxy for the detainees' moral character, which in turn may be perceived as the most reliable predictor of dangerousness and flight risk. In this way, certain heuristics may be playing an important role in providing cognitive shortcuts for immigration judges. These microlevel mental processes, though difficult to study in natural settings, are ripe topics for future research.
Another possible explanation for the exclusive significance of criminal history in bond decisions relates to a certain type of risk-aversion that might be prevalent among immigration judges. Given that immigration judges may be removed from the bench, the desire to not “rock the boat” may be common (Reference MarksMarks 2012:29). In this context, immigration judges may be especially motivated to avoid erroneous bond decisions involving detainees with certain types of convictions. As one criminal judge explained in describing pretrial custody decisions, “If you let [the defendant] out … and then the victim was badly injured, or killed, you have the problem of the newspapers coming in a very critical vein” (Reference SuffetSuffet 1966:330). To the extent immigration judges face similar pressures, immigration judges might find it less professionally “risky” to focus on prior criminal histories and to err on the side of detaining noncitizens with multiple or serious criminal convictions.
In addition, certain contextual factors in immigration bond hearings might reinforce perceptions of criminality of immigrant detainees. Although immigration detention is legally defined as administrative rather than criminal in nature, immigrant detainees are required to wear government-issued uniforms and wristbands with identifying information at all times, including at their bond hearings. These uniforms are color-coded based on ICE's assessment of security risk. In terms of their general appearance, there is little to no difference between the uniforms worn by immigrant detainees and those worn by criminal inmates. Moreover, immigrant detainees are routinely shackled when they are transported from the detention facilities to the courthouses for in-person hearings, and they remain restrained during their hearings. In short, physical markers of criminality imposed on the detainees during their bond hearings may serve as a powerful anchor for immigration judges. These physical markers of criminality may also make the existing societal tropes about the purported “lawlessness” of unauthorized immigrants (see Reference RyoRyo 2013, Reference Ryo2015) highly salient for the immigration judges.
Finally, it is instructive to consider the broader contemporary trends in immigration law and policy that might influence—directly or indirectly—the decisional environments of immigration judges. Over the past few decades, the growing scope and the reach of crimmigration law have increasingly blurred the boundaries between immigration and crime control (see, e.g., Reference Beckett and EvansBeckett and Evans 2015; Reference LightLight 2014). Underlying this trend is the movement of immigration law toward the “crime control model”—one of the two classic models that Reference PackerPacker (1968) originally developed to describe the workings of the criminal process. Unlike the due process model, which stresses adherence to procedure and the protection of individual rights in the face of state power, the crime control model holds that the efficient and swift “repression of criminal conduct is by far the most important function” of the criminal process (Reference PackerPacker 1968:158).
Crimmigration law is a culmination of immigration law's embrace of this crime control model in two important respects. First, as Reference Bosworth and KaufmanBosworth and Kaufman (2011:431) have noted, with the expanding web of crimmigration law, noncitizens have become the “next and the newest enemy” in the U.S. war on crime. Second, crimmigration law emphasizes efficient modes of crime control by privileging “the moment of the crime as the determining factor for often-permanent expulsion” from the United States (Reference StumpfStumpf 2011:1709). Viewed against this broader backdrop, the exclusive significance of criminal history in bond decisions might be a reflection of the extent to which the crime control imperatives of crimmigration law have become the operating norm in immigration courts. Further research is needed to understand how and under what conditions bond decisions may not only reflect, but also reinforce and perpetuate, the operation of this norm in immigration proceedings beyond bond hearings.
Immigration laws are currently silent on the relative importance of various legally relevant factors and thus provide no guidance on how each factor should be weighed and integrated. Furthermore, immigration judges are not provided basic information about whether the detainees that they have released on bond have returned to court and/or have committed other offenses on release. Given these gaps in legal and informational cues, some observers might argue that immigration judges are acting “rationally” (in the neoclassical utility-maximizing sense) by deferring heavily to decisions made by judges in the criminal justice system. Yet from the perspective of individual detainees, these bond decisions may appear arbitrary, largely shaped by the personal whims of individual judges. For example, Edson, one of the released detainees whom I interviewed, described the bond hearing outcomes this way: “It's just the luck you have, honestly … The judge you get and how they're feeling that day.”Footnote 18 Whether and to what extent immigration judges ought to be given more concrete legal guidance on how to weigh and integrate the legally relevant factors, and how to bring greater transparency to their decisions-making process are key questions that have broad implications for the legitimacy of immigration courts.
Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's web site:
Table A1. Type of Criminal Convictions
Table A2. Characteristics of Respondents With versus Without Criminal Convictions
Table A3. Description of Measures Used in the Bivariate and Multivariate Analyses
Table A4. Operationalization of Legally Relevant Factors
