Representing Immigrants: The Role of Lawyers in Immigration Bond Hearings

Abstract

Do immigration lawyers matter, and if so, how? Drawing on a rich source of audio recording data, this study addresses these questions in the context of U.S. immigration bond hearings—a critical stage in the removal process for noncitizens who have been apprehended by U.S. immigration officials. First, my regression analysis using a matched sample of legally represented and unrepresented detainees shows that represented detainees have significantly higher odds of being granted bond. Second, I explore whether legal representation affects judicial efficiency and find no evidence of such a relationship. Third, I examine procedural and substantive differences between represented and unrepresented hearings. My analysis shows no differences in the judges' procedural behaviors, but significant differences in the detainees' level and type of courtroom advocacy. Represented detainees are more likely to submit documents, to present affirmative arguments for release, and to offer legally relevant arguments. Surprisingly, however, I find no evidence that these activities explain the positive effect of legal representation on hearing outcomes. These findings underscore the need to investigate not only what lawyers do in the courtroom, but also less quantifiable factors such as the quality of their advocacy, the nature of their relationship to other courtroom actors, and the potential signaling function of their presence in the courtroom.

Do lawyers matter, and if so, how? These are core questions of longstanding interest to scholars, legal advocates, and policymakers alike across many different areas of law (see, e.g., Eagly and Shafer 2015; Quintanilla et al. 2017; Shanahan et al. 2016a; Taylor Poppe and Rachlinski 2016). Because lawyers play such a central role in the American legal system, the answers to these questions have substantial implications for access to justice, inequality, and the rule of law in the United States. This study presents new data and a novel approach to addressing these questions in the context of U.S. immigration law.

The specific empirical focus of this study is immigration bond hearings. In these hearings, immigration judges must determine whether noncitizens should be released or continue to be detained pending the completion of their removal proceedings.¹ Given the personal liberty interests at stake, bond hearings constitute a critical stage in the removal process that can have deep and lasting social, economic, and legal consequences for the noncitizens and their families (Ryo 2016). In addition, immigration bond hearings can function as an “informal discovery tool” through which noncitizens can obtain the records, affidavits, and other evidentiary materials needed in their removal defense.²

Understanding the role of lawyers in removal proceedings in general, and bond hearings in particular, is a timely and urgent task in light of the recent trends in U.S. immigration enforcement. The total number of removals increased significantly under the Obama administration, which had inherited a “formidable immigration [enforcement] machinery” (Chishti et al. 2017). More recently, Immigration and Customs Enforcement (ICE) reported a sharp increase in the total number of ICE interior removals under the Trump administration's mass detention and deportation policy (U.S. Department of Homeland Security 2017). Many noncitizens in removal proceedings are not legally represented. A recent national study of over 1.2 million immigration removal cases decided between 2007 and 2012 found that 63 percent of all individuals and 86 percent of detained individuals lacked legal representation (Eagly and Shafer 2015). This pattern is not surprising. Individuals in removal proceedings, unlike defendants in criminal trials, are not provided government appointed legal counsel because removal proceedings are considered to be civil or administrative under the law.³

Yet, removal proceedings are widely recognized to be adversarial in nature, often characterized by severe power differentials and resource disparities between the parties (Adams 2010). For example, the government is always represented by a Department of Homeland Security (DHS) trial lawyer with training in immigration law, whereas prosecuted individuals are noncitizens who often lack English fluency, economic resources, and familiarity with our legal system. In short, many noncitizens are left to navigate on their own a complex body of law that one federal court described as resembling King Minos' labyrinth (Lok v. INS 1977: 38).

The belief that lawyers play a critical role in removal proceedings has been central to the creation of public funds in a growing number of cities to provide legal representation to detained noncitizens (see, e.g., Corser 2017; National Immigration Law Center 2016). Likewise, a belief in the importance of legal representation has been at the heart of high-profile class action lawsuits that have sought to secure particularly vulnerable subpopulations—such as noncitizens with mental disabilities, and unaccompanied children—the right to appointed counsel (Franco-Gonzalez v. Holder 2013; J.E.F.M. v Lynch 2016). The Supreme Court has also recognized that “the complexity of immigration procedures, and the enormity of the interests at stake, make legal representation in deportation proceedings especially important” (Ardestani v. INS 1991: 138).

Consistent with these commonly held beliefs about the importance of legal representation for noncitizens, empirical studies have documented a positive relationship between legal representation and favorable legal outcomes at various stages of the immigration court process (for a summary, see Eagly and Shafer 2015). The current study makes both empirical and theoretical contributions to this growing body of research and the broader literature on the role of lawyers in the civil justice system.

Empirically, the only way to capture certain micro-level information about what happens during the bond hearings (e.g., the hearing length, the behavior of the parties, and the rationale for the judicial decisions, etc.) is either through in-person courtroom observations, or through audio recordings of the hearings. While courtroom observations allow researchers to personally experience the hearings, this method presents significant data collection challenges (see Blanck 1987), especially in the context of bond hearings, which typically unfold rapidly within a short span of time. In contrast, audio recordings of the hearings offer researchers an intimate view of the hearings and also afford them an opportunity to systematically identify and code certain kinds of data (Ryo forthcoming). Audio recordings of court hearings thus constitute a valuable, yet often overlooked, resource in studies of legal representation.

Drawing on a unique set of audio recording and accompanying survey data, this study explores for the first time not only whether, but also how, legal representation might matter in immigration bond hearings. To undertake this exploration, the study develops a theoretically grounded approach for examining the various ways in which lawyers might matter in immigration bond hearings. The study's findings call into question conventional understandings about the role of lawyers in such hearings.

Background on Immigration Bond Hearings

In order to provide a context for understanding the role of lawyers in immigration bond hearings, I begin by offering an overview of the removal process in the United States. My discussion here is brief and is focused on those elements of the process that are the most relevant for this study.⁴ ICE may initiate a removal proceeding based on a noncitizen's violation of immigration laws or criminal convictions that render a noncitizen removable under the law (8 U.S.C. §§ 1182, 1227). The same removal process applies regardless of the noncitizens' residency status, and therefore applies equally to lawful permanent residents (LPRs).

Once ICE initiates a removal proceeding, the immigration judge must terminate the case if the government fails to state a valid ground for removal. If the case is not terminated, the noncitizen may seek legal relief from removal, such as asylum, cancellation of removal, and adjustment of status. If the immigration judge denies legal relief, the noncitizen will be ordered removed from the United States (8 U.S.C. § 1229a). The immigration judge's decision whether to grant 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 not part of the judicial branch of the federal government. Instead, they are under the jurisdiction of an executive agency within the Department of Justice known as the Executive Office for Immigration Review (EOIR). Immigration judges are attorneys appointed by the U.S. Attorney General to serve as civil servants (8 U.S.C. § 1101(b)(4)). Thus, immigration judges do not have the same degree of judicial independence as federal judges who derive their authority from Article III of the Constitution and have lifetime appointments. Another salient aspect of immigration courts and judges is that immigration judges generally face extraordinarily high caseloads, with little time and staff support to adjudicate those cases (Marouf 2011: 431-34). In addition, immigration judges' decisions are subject to relatively limited administrative and judicial review (Hausman 2016; Marouf 2011).

Prior or subsequent to the initiation of removal proceedings, ICE has authority to detain noncitizens under the “discretionary detention” or “mandatory detention” provisions of the Immigration and Nationality Act (INA). For noncitizens who are held under the discretionary detention provisions, ICE may release the noncitizen or continue to detain the noncitizen while his or her immigration case is pending (8 U.S.C. § 1226(a)). ICE's custody decision may be appealed to the immigration court, and the immigration court's decision may be appealed to the BIA.

Different procedures exist for noncitizens held under the INA's mandatory detention provisions. These individuals include, for example, (1) certain classes of “arriving aliens,” including asylum seekers who have not yet passed their credible fear determination, and (2) noncitizens convicted of certain crimes enumerated in the INA (National Lawyers Guild 2017). These detainees are generally ineligible for release except under narrowly defined circumstances. Some federal courts, however, have held that prolonged detention entitles the detainee to a bond hearing before an immigration judge (see Baldini-Potermin 2016). Most recently, the Ninth Circuit Court of Appeals in Rodriguez v. Robbins affirmed such a right for noncitizens who are continuously detained for 180 days or more (see Rodriguez v. Robbins 2015).

Rodriguez is a class action lawsuit brought by long-term detainees in the Central District of California, and class members include mandatory detainees. Under Rodriguez, the immigration judges are required to release the detainees “on reasonable conditions of supervision … 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 2015: 1066). In February of 2018, however, the Supreme Court reversed the lower court's ruling on statutory grounds and remanded the case for further proceedings to determine whether the detainees have a constitutional right to periodic bond hearings (Jennings v. Rodriguez 2018).

In an immigration bond hearing (as in a criminal bail hearing), the judge must decide whether the noncitizen constitutes a danger to the community, and whether he or she poses a flight risk (In re Guerra 2006). In general, an 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). There are a number of legally relevant factors that immigration judges may consider in immigration bond hearings: the noncitizen's (1) possession or lack of a fixed address in the U.S., (2) length of residence in the U.S., (3) family ties in the U.S., particularly to those who can confer immigration benefits on the noncitizen, (4) employment history in the U.S., 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 n.d.: 6–7).

Existing Research and Theoretical Framework

The framework for my empirical analysis is informed by two related, but distinct, literatures: research on legal representation in criminal bail hearings, and research on legal representation in civil proceedings.⁵

Role of Lawyers in Criminal Bail Hearings

The Sixth Amendment right to appointed counsel applies to any “critical stage” of a criminal trial (United States v. Wade 1967: 224-25). However, whether bail hearings constitute such a critical stage remains an open question, and defendants in many states appear without legal counsel at their bail hearings (Bunin 2016). Against this legal background, researchers have focused their empirical investigations of legal representation in bail hearings on two main questions. First, scholars have asked whether legal representation matters at all for pretrial outcomes. Studies addressing this question generally do not distinguish between different types of legal counsel involved in the pretrial hearings. Thus, these studies focus on the average effect of (any type of) legal representation on case outcomes. Second, scholars have asked whether representation by privately retained counsel is associated with more favorable outcomes than representation by public defenders. On this question, study findings have been mixed (see, e.g., Turner and Johnson 2007; Williams 2017).

On the first question of whether legal representation (of any type) matters for pretrial outcomes, surprisingly little empirical research exists (see National Right to Counsel Committee 2015 for a review). The best-known contemporary study of this issue comes from the Baltimore City Lawyers at Bail Project (LAB) led by Douglas Colbert. Colbert and his colleagues conducted a randomized experiment to measure the effect of legal representation on the bail hearing outcomes of low-income defendants accused of nonviolent offenses (Colbert et al. 2002). The LAB study found that legal representation produced a host of “objective benefits.” For example, represented defendants were more likely to be released on their own recognizance and to receive lower bail amounts. In discussing why legal representation might have mattered for their study participants, Colbert and colleagues argued: “One reason is that represented defendants could better present beneficial and verified information … that supplemented the information provided by the pretrial release representative” (Colbert et al. 2002: 1755). This argument emphasizes the importance of lawyers' substantive expertise—that is, knowledge of what information is relevant and advantageous to present to the judge, and how to present such information.

Colbert and colleagues also examined whether legal representation was associated with “subjective benefits” (Colbert et al. 2002: 1748). To evaluate these subjective benefits, the LAB study surveyed the defendants' perceptions of system legitimacy and procedural fairness. An analysis of these survey responses indicated that represented defendants were more likely to believe that they were treated fairly and with respect by legal authorities and were more likely to express an intention to comply with the bail decision. Taken together, these findings suggest that legal representation improved the defendants' confidence in and satisfaction with the criminal justice system. Colbert and colleagues attributed these results in part to the expanded opportunities that represented defendants have “to tell their side of the story” or to be heard in court, which in turn promotes their belief that legal authorities have treated them as valued members of the community (Colbert et al. 2002: 1745).

Briefly summarized, the LAB study is notable in at least two respects. First, the LAB study used a randomized experiment (rarely seen in this area, as I discuss below) to address causal identification issues. Causal identification issues refer to problems related to separating observed associations between independent variables and outcome variables of interest into their causal and spurious components (Elwert and Winship 2014: 33). The LAB study is also notable for its focus on subjective or perceptual measures of procedural justice. To the extent that scholars of legal representation in the civil justice context have been concerned with the procedural impacts of legal representation, the focus typically has been on objective, rather than perceptual, measures of procedural justice.

Role of Lawyers in Civil Proceedings

The research on legal representation in civil proceedings has long acknowledged the challenges associated with identifying the causal effects of legal representation. One potential identification issue in this research arises from problems of selection bias. For example, if individuals with stronger claims are more likely to hire lawyers, or if lawyers systematically select cases that they are more likely to win, an observed relationship between legal representation and case outcomes would not necessarily mean that representation improved case outcomes. Another potential identification issue in this research relates to a problem commonly known as omitted variable bias. In this context, omitted variable bias refers to the possibility that both legal representation and case outcomes might be correlated with another factor. For example, if legal representation and English proficiency are positively correlated, and English proficiency has an independent effect on case outcomes, then a regression that omits English speaking ability will overstate the effect of representation on case outcomes.

An experimental design whereby legal representation is randomly assigned to cases can address these types of identification issues.⁶ For a number of reasons, however, randomized trials are rare in research on the U.S. legal profession. In fact, they are so rare that James Greiner and Andrea Matthews described each such trial as a “unicorn, a magical creation with no origin story that appears briefly in a larger setting and then fades away” (Greiner and Matthews 2016: 297). Despite the challenges facing empirical studies of legal representation, Emily Taylor Poppe and Jeffrey Rachlinski's recent review of research on legal representation in civil disputes has concluded: “On the whole … while there may be areas where legal representation is likely to have less of an impact on case outcomes, the bulk of the evidence indicates that lawyers matter” (Taylor Poppe and Rachlinski 2016: 942; see also Engler 2010).

Research on legal representation in civil proceedings has largely focused on the effect of representation on case outcomes (see, e.g., Nessel and Anello 2016; Ramji-Nogales et al. 2007; Srikantiah et al. 2015; Steering Committee of the New York Immigrant Representation Study Report 2011). However, at least one study has also examined the effect of representation on overall system efficiency (Eagly and Shafer 2015). Eagly and Shafer found in their national study of access to counsel that represented noncitizens were more likely to have their cases terminated, more likely to seek relief from removal, and more likely to obtain relief. But in addition to analyzing these case outcomes, Eagly and Shafer also examined whether legal representation produced measurable gains or losses in terms of system efficiency. Briefly summarized, their analysis showed that represented removal proceedings were associated with: (1) less time spent on seeking continuances to search for counsel, (2) reduced detention costs resulting from higher rates of bond release, and (3) lower rates of failure to appear at subsequent hearings (Eagly and Shafer 2015: 59–75).

Empirical studies that examine how lawyers might matter are also rare. As Catherine Albiston and Rebecca Sandefur have noted, these mechanism questions constitute an “unspecified black box” in many legal representation studies (Albiston and Sandefur 2013: 107). Sandefur's recent study of legal representation in a diverse array of civil proceedings (excluding immigration cases) represents one of the most comprehensive efforts to move the focus beyond final case outcomes to consider the ways in which lawyers might matter. Sandefur conducted a meta-analysis of 17 published studies involving more than 18,000 adjudicated civil cases across wide-ranging fields of law such as landlord/tenant, tax, employment, and social security disability insurance. Sandefur found that the lawyers' impact on case outcomes appeared to be the greatest in procedurally complex cases, whereas their impact was relatively marginal in substantively complex cases. In a related vein, Sandefur also concluded that one of the most important roles that lawyers played in the types of cases studied is that they helped courts to follow their own procedural rules (Sandefur 2015: 17).

Taken together, my review of the literature on legal representation in bail hearings and in civil proceedings suggests that legal representation might be consequential for both case outcomes and adjudication efficiency. The foregoing discussion of the existing literature also suggests that, to the extent that legal representation matters for case outcomes, it might matter in at least two distinct ways. Procedurally, lawyers might produce more favorable outcomes by forcing judges to adhere more strictly to the procedures established to ensure a fair adjudication. Substantively, lawyers might produce more favorable outcomes by providing zealous legal advocacy on behalf of their clients (e.g., providing relevant records to the court, presenting legally relevant arguments to the judges, etc.). I now explore each of these possibilities in detail in the context of immigration bond hearings.

Data and Methods

Data

The data for this study comes from two sources. The first dataset comes from the Rodriguez Survey. The Rodriguez Survey is an in-person survey of long-term immigrant detainees in the Central District of California who received a bond hearing notice pursuant to Rodriguez v. Robbins, the class action litigation described above (for additional details on the Rodriguez Survey, see Ryo 2016). Between May 2013 and March 2014, 565 detainees who were 18 years of age or older participated in the in-person survey. The survey was conducted as soon as practicable after the detainees' scheduled bond hearings; as a result, all but 36 detainees (6 percent) had a substantive bond hearing at the time of the survey. The survey captured diverse information, including: (1) detainees' demographic and case backgrounds, (2) pre-detention criminal and employment histories, (3) detention experiences, and (4) views about the law and legal authorities.

At the time of the survey, the detainees were held in four facilities across the Central District of California pending their removal proceedings. These facilities are the James A. Musick Facility (Musick), Theo Lacy Facility (Theo Lacy), Santa Ana City Jail (Santa Ana), and Adelanto Detention Facility (Adelanto). Approximately 23 percent of the respondents were held at Musick, 21 percent at Theo Lacy, 13 percent at Santa Ana, and 43 percent at 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. Adelanto is operated by a private prison company called the GEO Group, and it houses only immigrant detainees. At the time of the survey, ICE maintained contracts with each of these facilities to confine immigrant detainees pending their removal proceedings.

The second dataset comes from audio recordings of the bond hearings of a subset of the Rodriguez Survey respondents (Rodriguez Audio Data). To my knowledge, the Rodriguez Audio Data is the first and only dataset that systematically extracts information from audio-recorded hearings in immigration courts. Working with a team of law students, I coded the audio recordings for a variety of items across broad-ranging topics, including: (1) background information on the detainees, (2) basic information on the bond hearing, such as whether a witness testified, (3) the duration and outcome of the bond hearing, and (4) the discussion topics and the nature of exchanges between various actors in the courtroom, including the immigration judges, government attorneys, detainees, and the detainees' attorneys (to the extent that the detainees were legally represented).

For my analysis, I merged the Rodriguez Survey with the Rodriguez Audio Data. The merged dataset contains 430 hearings. About 29 percent of detainees in the merged dataset were LPRs and 67 percent were undocumented. About 4 percent of the detainees had some other legal status or unknown legal status. Given that there were only 29 female detainees in the merged dataset, I restrict my analysis to male detainees. Issues relating to the type of attorneys (e.g., legal services, law school clinical programs, private practice, etc.) and the quality of individual attorneys are beyond the scope of this study due to the nature of my data.⁷ Nonetheless, lawyer type and capability are important topics of inquiry that warrant further study, especially in light of the existing evidence that suggests that a lack of competent counsel may be a significant problem in removal proceedings generally (see, e.g., Posner and Yoon 2010; Schoenholtz and Bernstein 2008).

Before describing the variables used in my analysis, I briefly highlight the ways in which detainees in this study might not be representative of the immigrant detainee population nationally. Whereas all Rodriguez class members are contesting their removability and/or seeking legal relief from removal,⁸ that is not the case for all short-term detainees. Those detainees who do not contest their removability and/or do not seek legal relief from removal experience relatively shorter detention for the simple reason that they are removed from the United States. Further, Rodriguez class members may be more likely to have criminal convictions compared to short-term detainees, as some of the former were mandatorily detained due to their statutorily enumerated criminal offenses. Finally, given the substantial liberty interests implicated in long-term detention, Rodriguez class members are entitled to certain additional procedural protections that are not afforded to short-term detainees in their bond hearings.

Analytical Strategy and Measures

Supporting Information Table A1 contains detailed descriptions of all of the measures discussed below. My analysis proceeds in three stages. First, using a matched sample, I test whether legal representation is associated with higher odds of being granted bond (“matching analysis”). I preprocess the data using coarsened exact matching (CEM) to generate two groups of detainees who are comparable on various background characteristics but differ on their represented/unrepresented status. The premise underlying CEM, and any matching technique, is to approximate randomized trials as much as possible by pairing observations that are similar or identical (on relevant “pretreatment” covariates) but for the “treatment” condition (Stuart 2010).

CEM allows exact matching of comparison groups across multiple characteristics of interest. This preprocessing of the data produces a smaller sample for analysis, as unmatched observations are discarded. I then use regression adjustment to “clean up” any residual covariate imbalance between the groups (Stuart 2010: 13). The technical details of CEM and its advantages over other matching techniques, including propensity score matching, are well documented elsewhere (see, e.g., Blackwell et al. 2009; Iacus et al. 2012; King and Nielsen 2016). I use the cem routine in Stata to produce matched samples that differ on legal representation (yes/no), but are balanced on the following key covariates: Age (four groups based on quartiles: 18–29, 30–35, 36–43, 44+), English Speaking (two groups), Hispanic or Latino/a Origin (two groups), High School Degree or Higher (two groups), and Number of Felony Convictions (three groups: 0, 1, 2 or more).⁹ Finally, I estimate a logistic regression model of bond hearing outcomes (grant/deny) using the matched sample.

Second, to explore whether legal representation is associated with judicial efficiency (“efficiency analysis”), I adopt the common analytical approach used in existing studies of judicial decision making, which usually conceptualize efficiency in terms of case disposition time (see, e.g., Cauthen and Latzer 2008; Christensen and Szmer 2012). Thus, I analyze the relationship between legal representation and the following measures: (1) total hearing duration (in minutes), (2) whether any recesses occurred during the hearing, (3) whether the government reserved appeal, and (4) whether the detainee reserved appeal.¹⁰ I use the parties' reservation of appeal as a rough proxy for whether the parties actually appealed the judge's bond decision, as information on actual appeals are not available. Since appeals lengthen the overall case disposition time, the parties' decisions to appeal constitute one relevant measure of judicial efficiency (regardless of whether the appeals are meritorious or not).

Third, I compare the hearings in which the detainees were represented (represented hearings) and the hearings in which the detainees were pro se (unrepresented hearings) along various procedural and substantive dimensions (“procedural and substantive analysis”). The procedural dimensions I analyze include whether the judge: (1) discussed the detainee's eligibility for the bond hearing, (2) explained the burden of proof, (3) explained the standard of proof, and (4) informed the parties of their right to appeal the judge's decision. There are, of course, procedural decisions that judges may render outside of the hearings themselves, but such decisions are likely to be rare in the immigration bond hearing context. The substantive dimensions I analyze relate to various measures of courtroom advocacy, including whether: (1) any witnesses spoke during the hearing, (2) the government submitted any documents, (3) the detainee submitted any documents, and (4) the detainee and/or detainee's counsel made any affirmative arguments for release or conditions of release, and if such arguments were made, the number and type of arguments made.

In my analysis sample, the detainees and/or their counsel made affirmative arguments for release or conditions of release in a total of 241 bond hearings. I code these affirmative arguments by classifying them into 20 mutually exclusive categories. I describe each of these argument categories in detail in Supporting Information Table A2. Among the detainees who made an affirmative argument, the total number of arguments ranged from 1 to 10 (Mean = 3.51; SD = 1.83). The “conditions of release” refers to bond, monitoring, or other forms of supervision. In coding the affirmative arguments, I exclude the instances in which the detainee provided only a “yes” or “no” response to a line of questioning by the immigration judge or the government attorney. For detainees who were represented by counsel, affirmative arguments may have been made by the detainee or his or her counsel. The nature of the affirmative arguments was diverse, ranging from explanations of why a given detainee may have engaged in a criminal offense that led to his or her ICE custody, to discussions of hardship imposed on family members due to his or her continued detention.

Finally, for the measures in the procedural and substantive analyses that had significantly different means or proportions across represented and unrepresented hearings, I conducted a mediation analysis to assess whether those measures mediated the effect of legal representation on the odds of being granted bond. This analysis involves multivariate logistic regressions using the matched sample discussed earlier. A mediator is “a variable that explains the relationship between a predictor and an outcome” (Frazier et al. 2004: 116; see also Baron and Kenny 1986; MacKinnon et al. 2007).

Results

Descriptive Analysis

Approximately 60 percent of detainees in the analysis sample were granted bond. This grant rate, however, varies significantly depending on whether the detainees had legal representation. Specifically, 71 percent of represented detainees were granted bond. By contrast, only 48 percent of unrepresented detainees were granted bond. Table 1 provides basic descriptive statistics on the independent variables used in my matching analysis. As shown in Table 1, about 50 percent of detainees in the analysis sample had legal representation at their bond hearings.¹¹ The average age of the detainees at the time of their bond hearings was about 37 years old. About 54 percent of detainees reported that they spoke English very well/pretty well. The majority of detainees (88 percent) self-identified as Hispanic or Latino/a. About 44 percent of detainees had a high school degree or higher. On average, the detainees had 0.39 felony convictions, with about 72 percent of detainees having no felony convictions (not shown in Table 1).

Table 1. Means, Standard Deviations, and Correlations

Notes: N = 380 (after listwise deletion).

Matching Analysis: Representation and Hearing Outcomes

I generated a matched sample using the covariates discussed earlier. L1 is an index of the degree of global imbalance across the covariates. A value of 0 on L1 indicates perfect balance between comparison groups of interest (here, represented versus unrepresented detainees); a value of 1 on L1 indicates that no overlap exists between the two groups. As shown in Table 2, this multivariate L1 is significantly reduced post matching (from 0.43 to 0.23, a 47 percent reduction), indicating a substantial improvement in the overall balance of the sample. Using this matched sample, I re-estimated the original multivariate regression model shown in Table 2. The results of the regression analysis using the matched sample confirm that detainees with representation are significantly more likely to be granted bond than detainees without representation. More specifically, the odds of being granted bond are about three times higher for detainees with attorneys.

Table 2. Results from Logistic Regression Models of Attorney Effect on the Likelihood of Being Granted Bond, Nonmatched and Matched Samples

Notes: Each of the regression models includes the following covariates: Age (years), English Speaking, Hispanic or Latino/a, High School Degree or Higher, and Number of Felony Convictions. Cluster-adjusted standard errors in parentheses (adjusted for clustering at the judge level).

^*p < .05; ^**p < .01; ^***p < .001 (two-tailed tests).

Of note, I conducted supplemental analyses to determine whether judge characteristics should be included in the regression models. To conduct these supplemental analyses, I first coded the judges for their gender, the political party of the appointing U.S. Attorney General (Democrat or Republican), and their prior work experiences. No reliable data exist on the judges' race and ethnicity. The measurements capturing the judges' prior work experiences consisted of a series of indicator variables for whether the judge had ever worked for the government, the DHS, nongovernmental organizations, and private law firms. I examined the bivariate relationship between each of these judge characteristics and bond grant/deny decisions but found no statistically significant results. Given these results, I did not include judge characteristics in the regression models.

Efficiency Analysis: Representation and Judicial Efficiency

Are represented hearings associated with greater judicial efficiency? Table 3 shows that represented and unrepresented hearings do not differ on any of the efficiency measures I analyzed. The represented and unrepresented hearings lasted on average about 20 minutes and 18 minutes, respectively. Recesses are relatively uncommon in both represented and unrepresented hearings (19 percent and 20 percent, respectively). The government reserved appeal in about 18 percent of represented hearings, as compared to 13 percent of unrepresented hearings, but the difference is not statistically significant. The detainees were more likely to reserve appeal in represented hearings than in unrepresented hearings (82 percent and 79 percent, respectively), but again, this difference is not statistically significant.

Table 3. Descriptive and Bivariate Statistics for the Hearing Analysis

Notes: All variables are binary except for Hearing Duration (minutes) and Number of Affirmative Arguments. ^aOnly the hearings in which the judge granted bond (N = 240) were considered in generating the descriptive and bivariate statistics for Government Reserved Appeal; likewise, only the hearings in which the judge denied bond (N = 152) were considered in generating the descriptive and bivariate statistics for Detainee Reserved Appeal. ^bOnly the hearings in which detainees made affirmative arguments for release (N = 241) were included in generating the descriptive and bivariate statistics for Number of Affirmative Arguments.

^*p < .05; ^**p < .01; ^***p < .001 (two-tailed tests).

Procedural Analysis: Judges' Procedural Behavior

Next, I turn to the results of my procedural analysis. As shown in Table 3, I did not find significant differences in any of the procedural measures. The judges were equally likely to discuss the eligibility for bond in represented and unrepresented hearings (43 percent and 39 percent, respectively). In only a minority of cases did the judges explain the burden of proof: 19 percent and 17 percent, respectively, for represented and unrepresented hearings. The proportion of cases in which the judges explained the standard of proof was even lower: 7 percent and 8 percent, respectively, for represented and unrepresented hearings. Although judges were much more likely to inform the parties of their right to appeal (than to inform them of the burden and the standard of proof), the differences in these proportions across represented and unrepresented hearings were not significant (94 percent and 96 percent, respectively). Taken together, I find no evidence that legal representation impacted the judges' procedural behavior.

Substantive Analysis: Courtroom Advocacy

Finally, I examined several substantive measures of courtroom advocacy. As shown in Table 3, the bivariate test results are significant for three of these substantive measures. Represented detainees are more likely to submit documents than unrepresented detainees (70 percent and 38 percent, respectively). Represented detainees are also more likely to make affirmative arguments (75 percent versus 46 percent, respectively), and present a greater number of affirmative arguments (approximately 4 arguments versus 3 arguments). To explore how represented and unrepresented hearings might vary in terms of the type of affirmative arguments presented, I examined differences in the proportion of detainees (among those who made at least one affirmative argument) making various kinds of affirmative arguments across represented and unrepresented hearings.¹² The results of this analysis are summarized in Table 4.

Table 4. Analysis of Affirmative Arguments, by Legal Representation Status

Notes: ^aEach variable is binary, indicating whether the detainee (and/or the detainee's counsel) made a given affirmative argument. ^bProportion refers to the proportion of detainees (among those who made at least one affirmative argument) who provided a given affirmative argument.

^*p < 0.05; ^**p < 0.01; ^***p < 0.001 (two-tailed tests).

As shown in Table 4, represented detainees made certain kinds of affirmative arguments at a significantly higher rate than unrepresented detainees. These affirmative arguments include: (1) U.S. Social Ties, (2) Immigration Case Status, (3) Nature of Criminal History, (4) U.S. Economic Ties, (5) Criminal Case Status, and (6) Financial Circumstances. Notably, the first five of these affirmative arguments pertain to each of the factors enumerated in the Immigration Judge Benchbook as legally relevant to bond determinations (EOIR n.d.). In contrast, only one type of affirmative argument was more common among unrepresented detainees—namely, Seeking Mercy/Fairness. This affirmative argument includes pleadings with the judge for forgiveness and appeals to the judge to treat like cases alike (e.g., “My fellow detainee who was in the same situation was released on a low bond”). In short, this particular affirmative argument can be characterized as appealing to the judges' sense of empathy and basic norms of distributive justice—factors that are not enumerated as legally relevant for bond determinations (at least not formally).

Given that represented detainees are more likely to submit documents, to make affirmative arguments for release, and to present legally relevant arguments, I examined whether these variables might be mediating the relationship between legal representation and hearing outcomes. For there to be mediation, the explanatory variable (here, whether or not the detainee had an attorney at the bond hearing) must be significantly related to both the outcome variable (whether or not the detainee was granted bond) and the mediating variable (whether or not the detainee submitted documents, etc.). Moreover, the mediating variable must be significantly related to the outcome variable, controlling for the effect of the explanatory variable on the outcome variable. If mediation exists, the effect of the explanatory variable on the outcome variable will be significantly reduced when the mediating variable is added to the regression model (see Brader et al. 2008; Frazier et al. 2004).

Table 5 shows the results of a series of regression models from my mediation analysis, using the matched sample. Models 1a, 2a, 3a, and 4a examine the relationship between the odds of being granted bond and each of the following potential mediating variables: (1) the detainee submitted documents, (2) the detainee made an affirmative argument for release, (3) the number of affirmative arguments made, and (4) the number of legally relevant affirmative arguments made. Recall that the first three of these variables were significantly related to the legal-representation variable in the substantive analysis presented earlier. Given my earlier finding that detainees with representation are more likely to make legally affirmative arguments, I generated the fourth variable to capture the number of legally relevant affirmative arguments made.

Table 5. Results from Logistic Regression Analysis of Whether Substantive Measures Mediate the Attorney Effect on the Likelihood of Being Granted Bond, Matched Sample

Notes: N = 333. Each of the regression models includes the following covariates: Age (years), English Speaking, Hispanic or Latino/a, High School Degree or Higher, and Number of Felony Convictions. Cluster-adjusted standard errors in parentheses (adjusted for clustering at the judge level).

^*p < .05; ^**p < .01; ^***p < .001 (two-tailed tests).

The results of my mediation analyses are surprising. First, Models 1a, 2a, 3a, and 4a show that none of the potential mediating variables have a significant effect on the odds of being granted bond. Second, I added the attorney variable and a set of covariates representing the relevant detainee background characteristics in Models 1b, 2b, 3b, and 4b. These models show that none of the mediating variables diminish the effect of legal representation on the odds of being granted bond. In sum, contrary to the conventional wisdom that suggests that what lawyers do in the courtroom matters a great deal for their clients' case outcomes, I find no evidence that the level and type of courtroom advocacy explain the relationship between legal representation and favorable hearing outcomes.

Discussion and Conclusion

Drawing on a rich source of data and using a multi-pronged analytical approach, I investigated whether and how lawyers might matter in U.S. immigration bond hearings. First, to determine whether lawyers matter for hearing outcomes, I used a matching technique designed to mimic random assignment to produce a sample of represented and unrepresented detainees who are equivalent or comparable on relevant background characteristics. My regression analysis using this matched sample indicated that represented detainees have significantly higher odds of being granted bond than unrepresented detainees.¹³

Second, to determine whether lawyers matter for judicial efficiency, I analyzed the duration of the hearings, whether a recess was taken during the hearings, and whether the parties reserved appeal at the conclusion of the hearings. I found no significant differences between represented and unrepresented hearings on these efficiency measures. Third, I compared whether represented and unrepresented hearings differed along various procedural measures and substantive measures of courtroom advocacy. The results of this analysis indicated that whereas the judges' procedural behaviors did not vary significantly between represented and unrepresented hearings, there were significant differences in the substantive measures of courtroom advocacy across represented and unrepresented hearings. Most notably, represented detainees were more likely to submit documents, to present affirmative arguments for release, and to offer legally relevant arguments.

Surprisingly, however, I found no evidence that these differences in the observable courtroom activities mediated the relationship between legal representation and the odds of being granted bond. One possible explanation for this finding might be that my measures of courtroom advocacy do not capture the quality of advocacy. My data do not contain information about whether the documents submitted were relevant or helpful to the judges' bond determination. Nor do my data contain information about how relevant or helpful the judges found the detainees' affirmative arguments for release. To the extent that the overall quality of the submitted documents and of the affirmative arguments presented were poor (notwithstanding that such affirmative arguments were legally relevant), we might expect these particular measures of courtroom advocacy to have little explanatory power.

A number of existing studies provide some degree of support for this expectation. The immigration bar is often criticized as providing ineffective or low quality legal services (see Posner and Yoon 2010). But research suggests that the quality of services provided by immigration lawyers varies widely (Miller et al. 2015; see also Edwards 2017: 1503–05). Eagly and Shafer (2015: 52–54) find that small and solo law firms generally have the lowest level of success attaining case termination and legal relief for their clients, and nonprofits and law school clinics generally have the highest level of success. They also find that small firms and solo practitioners provided 90 percent of all removal representation (Eagly and Shafer 2015: 26). Based on these statistics, it is likely that small and solo law firms also account for the bulk of legal services provided in immigration bond hearings. These observations highlight the need for studies of representation quality and a comprehensive examination of measures of legal advocacy—both in and outside of the courtroom—in immigration bond hearings.

As “the absence of evidence is not evidence of absence” (see Greiner and Matthews 2016: 302), I urge caution in efforts to interpret the null findings. Nonetheless, insofar as the level and type of courtroom advocacy do not explain why legal representation is associated with more favorable bond hearing outcomes, two other possible explanations might warrant a close consideration. The first is the possibility that much of the lawyer action that contributes to favorable bond hearing outcomes involves what Rebecca Sandefur (2015) has referred to as lawyers' “relational expertise.” In contrast to “substantive expertise,” which relates to the lawyers' knowledge of the law (e.g., statutes, doctrine, case precedent) and legal procedures, relational expertise involves knowing how to navigate human relationships.¹⁴ At its core, relational expertise is “people knowledge … that guides [the lawyer's] interactions with judges, court staff, clients, and other attorneys” (Shanahan et al. 2016a: 490 (internal quotes deleted)).¹⁵

It is worth noting that relational expertise can take on two distinct forms. The first type of relational expertise arises from what Marc Galanter (1974) has called the repeat-player status of certain individuals or organizations in legal settings. Repeat players are persons or organizations “who are engaged in many similar litigations over time.” In contrast, one-shotters “have only occasional recourse to the courts” (Galanter 1974: 97). Some immigration judges and lawyers, by virtue of their repeated interactions with one another over time, likely come to develop a set of understandings and knowledge about each other's idiosyncrasies, preferences, and working styles (see Eagly 2015: 988–89).

The second type of relational expertise is not contingent on the lawyers' repeat-player status with respect to any given set of judges or government lawyers. Instead, this type of relational expertise arises from a more generalized set of “skill[s] at negotiating the interpersonal environments in which professional work takes place” (Sandefur 2015: 16). The following anecdote is illustrative of this type of relational expertise. K. Craig Dobson, a noncriminal lawyer, found himself having to represent a friend charged with driving under the influence (DUI). Not knowing where to begin, Dobson bought a book written by a leading DUI attorney. Dobson describes what transpired when he asked his attorney colleagues about a procedure described in the book (Dobson 2017: 1):

I asked one lawyer about the procedure that he used to test the equipment at the police station that measures blood alcohol content. The colleague laughed and said that nobody really did everything … recommended in his book… . [H]e said that this would likely just make some people mad, namely the judge and the prosecutor, and ultimately hurt not only this client, but also my reputation and thus future clients.

Both this type of relational expertise, and the relational expertise derived from the repeat-player status I have described above, likely generate distinct advantages to the lawyer's clients.¹⁶ But researchers cannot easily capture these kinds of dynamics through conventional measurements of courtroom activity. Consequently, existing studies of the role of immigration lawyers have yet to consider the relative importance of relational expertise in immigration proceedings.

That the level and type of courtroom advocacy may not explain the positive relationship between legal representation and hearing outcomes also raises the possibility that perhaps even more important than what lawyers do, their mere presence in the courtroom might serve an important signaling function that advantages their clients. According to a recent experimental study by Victor Quintanilla and colleagues (2017), employment claimants' pro se status triggered negative stereotypes about the claimants that negatively impacted their settlement awards. More specifically, Quintanilla and colleagues found that law-trained decision makers (law students and lawyers) perceived pro se claimants as less competent than legally represented claimants, and that these competence stereotypes mediated the effect of pro se status on settlement awards. If similar dynamics were operating in the immigration law context, we should expect detainees who appear in court with lawyers to obtain more favorable outcomes, independent of any courtroom advocacy in which the lawyers might engage. One former detainee with whom I interviewed hinted at such a dynamic in this way: “With a lawyer, the judges treat you differently—automatically.”¹⁷

Systematic investigations of the relative importance of lawyers' relational expertise and their possible signaling function promise a more complex and nuanced understanding of the role of lawyers in immigration bond hearings. I conclude with a discussion of some of the other limitations of this study that might serve as useful starting points for future research on legal representation in immigration proceedings more generally.

This study focused on the role of lawyers, but future studies should also examine the role of nonlawyer representatives and professionals in immigration proceedings (see, e.g., Shannon 2011; Thompson 2016). In the U.S. immigration law context, individuals accredited by the BIA and certain categories of persons who are expressly recognized by the immigration court can present cases in immigration court (EOIR 2017: 17). Some observers have noted that “[n]onlawyer representation of persons in removal proceedings exacerbates the likelihood of ineffective or incompetent representation” (Medina 2012: 472). But the question remains: With appropriate training and supervision, can nonlawyer representatives have the same or similar kind of impact as lawyers on case outcomes? Currently, no empirical data are available to adequately address this question.

An important related question is whether partial or “less than full representation” is better than no representation at all (see Shanahan et al. 2016b). For example, how does formal legal representation compare to pro se assistance in which noncitizens receive advice from lawyers in handling their own cases? These questions bring to the fore issues of relative cost, efficiency, and efficacy of different types of representation and assistance. To make informed policy decisions about how best to allocate limited resources for maximum impact in immigration courts, we need a better understanding of different models of legal services and varying approaches to regulating those services (Hadfield and Rhode 2016; Sandefur 2010).

In addition, future research should consider the longitudinal nature of immigration proceedings. The existing research does not provide a comprehensive understanding of whether and how lawyers might matter throughout the removal process, from the point of apprehension to the merits hearings. Is the impact of legal representation greatest at the early or later stages of the removal process? Do judges view lawyers and their performances differently across these different stages? Are noncitizens who retain different lawyers for different stages of the removal process (for example, one lawyer for the bond hearing and a different lawyer for the merits hearing) at a significant disadvantage compared to noncitizens who retain the same lawyer? These and related questions prompt us to take a more holistic view that attends to the longitudinal and interconnected nature of various stages in the removal process.

Finally, this study focuses only on the “objective” legal outcome (the grant or denial of a bond) of the immigration bond hearings. But as Colbert and colleagues (2002) have noted in their study of the role of legal representation in criminal bail hearings, “subjective” outcomes may be as equally important as the objective outcomes (see also Zimerman and Tyler 2010). As I discussed earlier, Colbert et al.'s study, for example, shows that defendants who had legal representation were more likely to feel that they had been treated fairly and respectfully during their hearings, and to express greater satisfaction with the legal system. Moreover, studies show that people's beliefs in the legitimacy of legal authority and trust in the legal system are central to their willingness to obey the law and to cooperate with legal authority (Sunshine and Tyler 2003; Tyler 2006), including immigration laws and immigration authorities (Ryo 2013, 2017a). From this standpoint, whether and to what extent our immigration system might be fostering legal cynicism among noncitizens by failing to provide appointed counsel are critical topics for future inquiry (see Ryo 2017b).

In the current era of increasing migration control and stepped-up enforcement in the United States, a rapidly expanding number of noncitizens will be forced to navigate our complex legal system. The issue of whether they will have to do so alone or with the assistance of competent legal advocates is of paramount public interest. As Judge M. Margaret McKeown of the Ninth Circuit Court of Appeals wrote in a recent case involving the issue of lack of legal representation for minors in removal proceedings (J.E.F.M. v. Lynch 2016: 1041): “To give meaning to ‘Equal Justice Under Law,’ the tag line engraved on the U.S. Supreme Court building, … the problem demands action now.” This admonition applies equally with respect to hundreds of thousands of adult noncitizens in removal proceedings who must struggle through the system with little to no legal knowledge and resources. Continued research not only on the degree to which, but on how, legal representation matters in immigration proceedings will play an important role in developing a fair and effective administration of justice in immigration law in the United States.