Immigration Judges and U.S. Asylum Policy 0812246608, 9780812246605

Citation preview

Immigration Judges and U.S. Asylum Policy

PENNSYLVANIA STUDIES IN HUMAN RIGHTS Bert B. Lockwood, Jr., Series Editor

IMMIGR ATION JUDGES AND U.S. ASYLUM POLICY

Banks Miller, Linda Camp Keith, and Jennifer S. Holmes

U N I V E R S I T Y O F P E N N S Y LVA N I A P R E S S PHIL ADELPHIA

Copyright ©  University of Pennsylvania Press All rights reserved. Except for brief quotations used for purposes of review or scholarly citation, none of this book may be reproduced in any form by any means without written permission from the publisher. Published by University of Pennsylvania Press Philadelphia, Pennsylvania - www.upenn.edu/pennpress Printed in the United States of America on acid-free paper           A Cataloging-in-Publication record is available from the Library of Congress ISBN ----

For Rachel and Cal —Banks For Adam and Summer —Linda For those fleeing persecution and in search of freedom —Jennifer

This page intentionally left blank

CONTENTS

Chapter 1. Introduction

1

Chapter 2. Creating a Dataset

26

Chapter 3. A Cognitive Approach to IJ Decision Making

48

Chapter 4. Local Conditions and IJ Decision Making

84

Chapter 5. Appealing to the Board of Immigration Appeals

106

Chapter 6. The Policy Gap and Asylum Outcomes

150

Chapter 7. IJs and Reform of the U.S. Asylum System

187

Notes

201

References

219

Index

233

Acknowledgments

239

This page intentionally left blank

CHAPTER 1

Introduction

In this book we seek to enhance understandings of why immigration judges (IJs) do what they do. We perceive IJs as the linchpin of U.S. asylum policy, and we assert in these pages that understanding how IJs decide asylum cases is the best place to begin trying to grasp asylum policy in the United States. In addition, the IJs offer an interesting case study from the perspective of scholars of judicial behavior because they decide cases in highly ideological fashion even though they are analogous to trial judges, a situation that is not often depicted in the literature. We attempt to move beyond the asylum literature’s focus on disparities in grant rates as the primary criticism of the U.S. asylum process. Instead we focus on theoretical constructs—largely adapted from theories of judicial decision making—that allow us to better understand the conditional nature of IJ decision making. Th is approach leaves us with the overriding sense that eliminating disparities in IJ adjudication is akin to tilting at windmills—the causes of variation are too deeply embedded and the contexts in which decisions are made are too varied and influential. Instead, we implicitly focus on the quality of IJ decisions—a focus we make explicit in our final chapter, where we offer several concrete suggestions for improving the quality of decision making by IJs. U.S. asylum policy represents a unique intersection of foreign and domestic policy. The implementation and adjudication of asylum cases triggers a wide range of potentially confl icting interests including international human rights norms, national security issues, geopolitical interests, border and immigration control, and the national and state economies. Asylum adjudications are enmeshed in a complicated web of domestic immigration law, U.S. treaty obligations, and federal jurisprudence. Recently, significant changes have been made in U.S. asylum law in response the terrorist attacks on the United States and in response to fears of economic migrants flooding

2

Chapter 1

the country. These changes have in turn triggered concerns that the law has become too draconian to meet our treaty or broader humanitarian obligations and ultimately may have left legitimate asylum seekers vulnerable. The individual adjudications of asylum law are made in an asylum system that spans two executive branches (the Department of Justice and the Department of Homeland Security) and that are overloaded and underresourced at every stage of the process. Components of the system have been politicized, such as the hiring process of IJs under Attorneys General Ashcroft and Gonzales and the ideological culling of the Board of Immigration Appeals by Ashcroft. The key adjudicators—IJs—have generated a significant amount of controversy in regard to the consistency and fairness of their judgments, to such an extent that some critics have concluded that the decision of whether an individual gets asylum depends mostly upon the judge the individual draws. Both IJs and the Board of Immigration Appeals (BIA) have drawn criticism from the federal courts for the quality of their work. IJs in turn point to their crushing caseloads, limited support, and complicated cases, and limited independence from the Department of Justice (DOJ)—a point repeatedly made to multiple presidents and to Congress—most recently appealing to both the Senate and the House in the current attempts at immigration reforms. These issues have generated increasing concern and concomitant scrutiny of the asylum process and its various actors by commentators, activists, and scholars. In this book we engage in a theoretically driven, systematic, and rigorous examination of the core asylum adjudicator—the IJ. Our study allows us to empirically test the key criticisms and issues raised in regard to asylum decisions and U.S. asylum policy more broadly. Our theoretical underpinning allows us to offer explanations of IJ decision making that may inform future asylum policy or reforms. In this book, we seek to better understand U.S. asylum policy by focusing on those whom we consider to be the most important, yet relatively unstudied, actor in the convoluted asylum bureaucracy: IJs. We have decided to focus on IJs for a host of reasons, some theory driven, some substantively driven, and others data driven. Most important from our perspective, IJs decide the majority of asylum cases in the United States and decide them with a significant degree of finality. The BIA, the body responsible for review of the decisions of IJs, reviews only 47 percent of the merit asylum decisions made by IJs (a high percentage to be sure) and upholds 74 percent of the decisions that they review. Combining these percentages means that 12 percent

Introduction

3

of the decisions made by IJs are reversed. Although this reversal rate may be high in comparison to the rates of Article III courts, nonetheless IJ decisions are usually final. Below the IJs in the asylum bureaucracy are the asylum officers (AOs). They are the first to review affirmative asylum cases (we explain the distinction between affirmative and defensive asylum claims below). AOs tend to act as a rather permeable fi lter in the process, culling the cases in which it is clear that the claimant has a right to asylum and then passing the more difficult cases up the chain to the IJs. About one-third of the applicants who are eventually granted asylum receive it from AOs at this early stage of the process. In short, IJs decide the majority of asylum claims and virtually all of the cases in which asylum is not clearly due to the applicant. For these reasons, IJs tend to decide substantively more difficult cases than do the AOs. Our interest is also in analyzing the individual decision making of the IJs as opposed to summaries of aggregate trends across courts or time, or comparisons within courts. This is because we believe, theoretically, that an understanding of the asylum bureaucracy starts with an understanding of the case-level decision-making process of IJs. That is, how do IJs decide each individual case before them? Which factors matter most and how? Thankfully, data are available on the decisions made by individual IJs in cases from at least 1990 onward. Such data are not publicly available, to our knowledge, for the AOs at all and, though individualized case outcomes are available for the BIA, data are not available on the decisions of individual members in the BIA.1 Therefore, given our theoretically driven focus on individualized decision making, it is not possible to study the AOs or the BIA in sufficient detail. We recognize that IJs make decisions at the midlevel in a complex bureaucracy, and that the context in which they make their decisions is crucial to understanding them. Therefore, we are able to utilize the aggregate statistics that we generated for both the AOs and the BIA where appropriate to ensure that we have properly contextualized the decision making of the IJs without also overcomplicating our desired focus. Furthermore, in Chapter 5 we focus on how the BIA affects IJ decision making. We undertake the most thorough examination of the decision making of IJs ever attempted in the literature. To that end we analyze the IJs from a number of perspectives, beginning with a basic cognitive model of their decision making in cases that incorporate the extensive literature on judicial decision making as well as the concerns of international relations scholars. We engage in a cross-sectional analysis of all of the decisions of IJs in every

4

Chapter 1

asylum claim decided on the merits between 1990 and 2010, a total of over half a million claims. Although Congress has limited what data are available about applicants due to privacy concerns, we do know the form of relief requested, case number and outcome, applicant’s country of origin and language spoken, and whether or not the applicant had a lawyer. From this, we are able to leverage country of origin characteristics to form a more complete picture of the applicant’s case for asylum. We then expand our perspective to focus on how local demographic, economic, and political factors affect IJ decision making, drawing upon dominant theories within the broader immigration literature including variations of contact and threat theory. We next account for how the prospect of BIA review might influence IJ decision making. Finally, we analyze the aggregate trends in asylum decision making across the span of our data in an attempt to better understand the dynamic nature of policy interventions in the asylum process, drawing upon key policy perspectives within the immigration literature. Our theoretically driven focus and rigorous empirical analysis allow us to generate a substantive understanding of asylum decision making that should speak directly to various policy makers considering a variety of proposals that have been advanced for reforming the asylum system and the broader immigration system. Critics argue that past policies, laws, and DOJ regulations promulgated in response to the threat of terrorists, undocumented immigrants, and an unwieldy asylum system have generated unintended consequences for both the asylum seeker and the U.S. asylum bureaucracy. We examine these expectations empirically and believe that these analyses can inform a discussion about the potential unintended consequences of currently debated reform. Indeed, this is one of the major benefits of our approach: because we have a well-developed theory of how the IJs decide cases, we have insight into which parts of that decision-making process are amenable to change and how policy makers might go about making the desired changes. We turn next to a brief summary of asylum law and the recent controversies that have surrounded IJs.

Asylum and Refugee Status in International Law Following World War II, in 1951 the United Nations General Assembly promulgated the Convention Relating to the Status of Refugees (hereafter the Convention), which sets out rights and obligations of states to refugees. The

Introduction

5

core principle of non-refoulement underpins the international refugee regime. Generally speaking, this principle prohibits states from forcibly returning individuals who fear a return to their country of origin. For an individual to be eligible for protection, this concern must be based on a well-founded fear of being persecuted for reasons of race, religion, nationality, membership of a particular social group, or political opinion.2 While the general intent of the regime is protection of refugees, from the outset the Convention reflected the Eurocentric political realities of the time and provided legal protection to only a limited set of refugees. To qualify, individuals had to be refugees as a “result of events occurring before 1 January 1951.” In addition, countries were able to exclude refugees from outside of Europe. To do so, they were allowed to make a “declaration when becoming party, according to which the words ‘events occurring before 1 January 1951’ are understood to mean ‘events occurring in Europe’ prior to that date” (United Nations High Commissioner for Refugees [UNHCR] n.d., 2). The United States, which had played a strong role in restricting the scope of international protection of refugees under the Convention, did not join the 1951 Convention. For U.S. policy makers at that time “the most important aspects of American refugee policy were maintaining international attention devoted to refugees from Communist countries, encouraging emigration from the Eastern bloc, and minimizing appeals for assistance funds to refugees” (Betts, Loescher, and Milner 2012, 21). Instead it created “two other U.S.-led organizations that were parallel to and outside the purview of the United Nations” (Loescher 2003, 7 and see also Copeland 2003, 108–9) to limit the functional scope of the UNHCR. Over time the legal norm expanded, in large part due to the efforts of the UNHCR, which successfully pushed for universal coverage for refugees, which was achieved with the promulgation of the 1967 Protocol eliminating the original geographical and temporal restrictions (Betts, Loescher, and Milner 2012, 30).3 The United States strongly supported and joined the 1967 Protocol.4 The move clearly signaled a change from its previous unilateral Cold War policies, which we discuss in the next section. Today the right to non-refoulement is widely considered to be a part of customary international law and thus would apply even to states that are not formally a party to the 1951 Convention or Protocol (Goodwin-Gill and McAdam 2007, 345 and citations within). In 2001, the state parties to the Convention issued a declaration “reaffi rming their commitment to the 1951 Convention and the 1967 Protocol” and “recognizing in par ticu lar that the

6

Chapter 1

core principle of non-refoulement is embedded in customary international law” (UNHCR n.d., 4).

U.S. Asylum and Refugee Law and Policy Asylum law in the United States is domestic law that is “expressly based on international law (Anker 2011, 2). As Cianciarulo notes, the passage of the 1980 Refugee Act, “which sought to give statutory meaning to our national commitment to human rights and humanitarian concerns . . . ushered in a new era of refugee protection” (Cianciarulo 2006, 109–10). The 1980 Refugee Act was the first congressional act to specifically address refugees and asylum seekers. Previously, Congress had controlled immigration through rigid quota systems such as those implemented under the Immigration and Nationality Act of 1952 (INA), although presidents spanning FDR to LBJ had creatively skirted congressional limits by directing their attorneys general to use their power of parole to allow aliens, and thus unilaterally had admitted large numbers of refugees outside of the quota systems.5 The 1980 Refugee Act repealed the INA’s geographical and political limitations and lifted numerical caps on the number of annual asylum grants (Cianciarulo 2006). Cianciarulo argues that with the passage of the 1980 Act (along with the Supreme Court’s recognition of the implications of the Act in INS v. Cardoza-Fonseca), “asylum was no longer an ad hoc, marginal immigration procedure entirely subject to the whims of policy” (110). Anker (2011, 17) argues that Congress’s intent in enacting the legislation was “to conform provisions of U.S. law to the Refugee Convention.” According to the House committee report, the act represented an intention to emphasize and make paramount “the plight of refugees themselves, as opposed to national origins or political considerations” (as cited in Gibney 1988, 111). However, this does not mean that it was entirely free of foreign policy or national security interests. Gibney (2004b, 157) argues that even after the 1980 law, “through State Department opinions, the ideological predilections of the Administration found their way into Immigration and Naturalization Ser vice (INS) asylum decisions” (see also Ferris 1987, 126). We discuss this more in Chapter 3 when we examine the role of U.S. foreign policy interests in asylum outcomes. Today, there are two paths through which individuals may claim refugee status in the United States—either as a refugee or as an asylee. Both paths

Introduction

7

require that individuals fulfill the definition of refugee in the INA. Refugees apply for status outside of the United States, while individuals requesting asylum do so from within the United States or upon arrival at a port of entry. Refugee quotas and regional allocations are set by the president, who consults with Congress. In addition to INA eligibility, refugees must come from a country that is of “special humanitarian concern to the United States” and must not be resettled in another country or ineligible due to security, criminal, or other factors as determined by U.S. Citizenship and Immigration Ser vices (USCIS) (Martin and Yankay 2012, 1–2). While asylum is discretionary under U.S. law, there are no numerical limits, as there are for refugees. The United States has been a major refugee receiving state. As Gibney (2004b, 132) notes, no Western state has admitted more refugees than has the United States since the end of World War II. However, refugee flows into the country have dramatically decreased, particularly since the September 11 attacks. Recently, the number of refugee admissions has subsequently begun a slight recovery, as seen in Figure 1.1, which plots the number of asylum grants by IJs compared to annual refugee inflow from 1990 to 2010. There is a striking negative relationship between the two series, which have (at the annual level) a correlation of –.91. Each series is plotted on its own scale, with the number of asylum cases on the right axis and refugee inflow on the left axis. Since these are the two paths through which individuals can escape persecution and seek to gain legal protection within the United States, it is reasonable to find that trends across the two potential routes to safety in the United States interact. Exploring this connection is beyond the scope of this work, but we plan to address the relationship in subsequent work that focuses on the broader U.S. refugee system.

The Asylum Process Today U.S. law provides three treaty-based forms of relief or protection for individuals fleeing persecution: (1) asylum and (2) withholding of removal— which are based on the 1951 U.N. Refugee Convention and the 1980 Refugee Act—and (3) protection against return under Article 3 of the Convention Against Torture (CAT). The CAT narrowly prohibits the return of a person to another country if there are substantial grounds to believe he or she would be subjected to torture. Since 1999, the United States has also been bound under this obligation through the Foreign Affairs Reform and Restructuring

Chapter 1

40000

5000

10000 Asylum Grants

Refugee Admissions 60000 80000

15000

100000

120000

20000

8

0

20000

r = –0.91

1990

1995

2000 Year Refugee

2005

2010

Asylum

Figure 1.1. Asylum Grants and Refugee Admissions, 1990–2010.

Act of 1998. Each of the three forms of protection offers different levels of relief or benefits and each has somewhat varying legal thresholds that must be met.6 We discuss these fully in Chapter 2. Next we turn to the asylum process and the institutions and actors within the various stages of the process. The asylum process today involves two executive departments—the DOJ and the Department of Homeland Security (DHS). Generally speaking, jurisdiction over the asylum process between the DOJ and DHS can be demarcated as follows: “DHS has jurisdiction over ‘border’ or credible fear interviews and first instance affirmative asylum applications (for persons who voluntarily apply before the institution of removal proceedings)” and “DOJ has jurisdiction over asylum applications determined in the course of removal proceedings, as well as over withholding of removal and applications for protection the Convention Against Torture” (Anker 2011, 12). In January 1983, the Executive Office for Immigration Review (EOIR) was created as a separate agency within the DOJ through an internal DOJ reorgani-

Introduction

9

zation that combined the BIA with the IJ function previously performed by the former Immigration and Naturalization Ser vice (INS), making the immigration courts independent of INS, the agency that at the time was charged with enforcement of U.S. immigration laws. The EOIR is charged with administering immigration courts nationwide. The EOIR, located in Falls Church, Virginia, is headed by a director who reports directly to the deputy attorney general. Within the EOIR, the Office of the Chief Immigration Judge provides overall program direction of fift y-nine immigration courts throughout the United States and has administrative supervision for approximately 260 IJs. This number has been slowly increasing over time, in an attempt to reduce the backlog of cases. The BIA serves as the highest administrative tribunal adjudicating immigration matters, and has responsibility for interpreting and applying immigration laws nationally. The BIA is constituted by a directive of the attorney general and is authorized to have as many as fifteen members who serve at the pleasure of the attorney general. The BIA has broad authority to review decisions of IJs and does so through a paper review process. IJs are administrative adjudicators who are formally appointed by the deputy attorney general; however, the EOIR and the Chief Immigration Judge handle their hiring. Current qualifications set by the attorney general require only that the candidates have seven years of prior legal experience. IJs arguably have less structural independence than federal judges and potentially less independence than administrative law judges. Nonetheless, they maintain a high degree of independence. The INA (Section 240) states that “in deciding the individual cases before them . . . IJs shall exercise their independent judgment and discretion.” IJs (originally known as Special Inquiry Officers) act as trial-level judges at this stage with asylum hearings being somewhat adversarial in process if the applicant has an attorney. The Asylum Officer Corps (AOC) was created as a part of the INS belatedly in October 1990—a decade after the 1980 Refugee Act and in response to a push for further reform of asylum in 1990 during the Bush administration; the AOC was “intended to ensure that political-asylum rulings are ‘fair and sensitive’ ” (Koehn 1991, 232). AOs were placed under the authority of the DHS by the Homeland Security Act of 2002. They are now housed within the new USCIS, which has responsibility for enforcing federal immigration laws and administering immigration and naturalization benefits. There

10

Chapter 1

are eight asylum offices across the country, with over three hundred AOs serving. A noncitizen who is physically present in the United States may seek asylum through either an affirmative or defensive process (see Figure 1.2). In the affirmative process, the applicant voluntarily identifies himself or herself through an application with the USCIS. The individual may or may not have valid status in the United States at the time of the application, but the application is not initiated during removal proceedings. In the affirmative process, once an application is filed, the applicant will receive notice to be fingerprinted and then will receive a notice to appear for an interview with an AO, who will review the application in a nonadversarial process in which the applicant must bring his or her own interpreter if desiring one. AO decisions are reviewed by a supervisor with the AOC. AOs are empowered to grant asylum, but the rates are rather low, as we can see in Table 1.1. The grant rate ranges from a low of 22 percent in 1997 to a high of 47 percent in 2000. As of 1995, if the application is denied and the applicant does not have valid immigration status, the applicant is referred to the immigration court for a de novo hearing and is now in removal proceedings.7 At this point the asylum seeker enters the defensive asylum process. The referral rates have ranged from 49 percent in 2000 to 73 percent in 1996. In the defensive process, typically the noncitizen has been apprehended within the United States and is in removal proceedings in immigration court when the applicant makes an application for asylum. A second stream of defensive applicants consists of aliens who arrive at a U.S. port of entry without proper documentation and who are placed in the expedited removal procedures that went into effect in 1998 under the Illegal Immigration Reform and Immigrant Responsibility Act of 1996 (IIRIRA), which we will discuss further in subsequent chapters. If these individuals express a fear of persecution, they are detained and receive a “credible fear” interview with an AO; otherwise the immigration officer at the port of entry can deny admission and summarily remove the aliens. If the aliens are found credible by the AO, the individuals are referred to an IJ for a hearing. In the defensive process, applicants can apply for all three forms of relief if appropriate. The EOIR provides the applicants with an interpreter for the IJ hearings, but representation is not provided. If the applicants are not represented, the IJ guides them through the proceedings and advises them of their rights and options for relief or removal (Alexander 2006; Transactional Record Access Center [TRAC] 2009). Judges usually provide an oral

Interview by Asylum Officer (AO)

Applicant files for asylum with USCIS (affirmative path)

AO grants asylum

Arrested for immigration violation by Department of Homeland Security (DHS) (defensive path)

AO denies asylum

Notice to appear/Master calendar hearing before Immigration Judge (IJ)

IJ denies relief to applicant

IJ grants relief to applicant

Merits hearing before IJ

Asylum applicant appeals IJ decision

Board of Immigration Appeals (BIA) decides appeal

BIA grants relief to applicant

DHS appeals IJ decision

BIA denies relief to applicant

Attorney General (AG) may vacate any BIA decision by appeal from DHS or own volition

Asylum applicant appeals decision

Federal Courts of Appeal (Circuit Courts)

Figure 1.2. Asylum Process. Adapted from “Flow Chart: Steps in the Asylum Process,” at http://www.immigrantjustice.org.

Table 1.1. Descriptive Analysis of Asylum Officers and Immigration Judges Decisions, 1990–2010 Asylum Officers

Immigration Judges

Year

Number of Cases Decided

Grant Rate

Referral Rate

Number of Cases Decided

Asylum Rate

Withholding Rate

Withholding (CAT) Rate

Any Form of Relief

1990 1991a 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

— — 13,800 34,270 46,386 82,947 75,802 65,264 50,585 44,279 53,414 68,754 71,025 49,797 40,688 38,468 41,508 44,705 38,393 32,133 31,182

— — 0.34 0.25 0.27 0.24 0.24 0.22 0.27 0.44 0.47 0.40 0.35 0.30 0.33 0.34 0.32 0.26 0.34 0.36 0.37

— — — — — 0.64 0.73 0.72 0.68 0.50 0.49 0.56 0.61 0.65 0.62 0.61 0.58 0.62 0.53 0.58 0.59

12,167 8,284 8,883 8,892 11,847 22,165 30,980 28,393 27,550 27,483 26,319 26,573 32,334 38,452 35,432 32,450 32,363 28,408 25,020 22,423 20,162

0.22 0.24 0.25 0.21 0.19 0.17 0.18 0.23 0.28 0.31 0.36 0.37 0.35 0.35 0.36 0.36 0.42 0.43 0.43 0.45 0.49

0.00 0.01 0.01 0.01 0.01 0.00 0.00 0.01 0.01 0.03 0.03 0.03 0.03 0.04 0.06 0.07 0.08 0.08 0.07 0.08 0.08

— — — — — — — — — 0.00 0.00 0.00 0.00 0.00 0.00 0.01 0.01 0.01 0.01 0.01 0.01

0.22 0.25 0.26 0.22 0.20 0.17 0.18 0.23 0.29 0.34 0.39 0.40 0.39 0.40 0.42 0.44 0.51 0.52 0.51 0.54 0.58

Sources: Data gathered by authors from USCIS and EOIR. a Incomplete year so excluded from annual analysis.

Introduction

13

decision at the end of each hearing due to caseload constraints. On the right side of Table 1.1 we report the IJ grant rates. In the last column we report the grant rates for any form of relief (including asylum, withholding of removal, or withholding of removal under the CAT), which range from a low of 17 percent in 1995 to a high of 58 percent in 2010. Of the three forms of relief, applications for asylum have consistently been the most successful, ranging from grant rates of 17 percent in 1995 to 49 percent in 2010. Withholding of removal has been granted considerably less often, with a high of 7 percent or 8 percent from 2005 forward, and relief under the CAT has rarely been granted thus far. Understanding these time trends is important in attempting to describe the current state of U.S. asylum policy, and we take up that analysis in subsequent chapters. Both applicants and the DHS may appeal IJ decisions to the BIA, but the government appeals few cases in practice. IJ decisions are subject to the reasonableness standard of appellate review, which requires that the BIA find that the IJ reached unreasonable conclusions in order to reverse a decision. The facts of a case are reviewed for clear error by the BIA, whereas application of the law is reviewed de novo. Table 1.2 presents the annual number and percentages of asylum cases appealed in between 1990 and 2010, along with the percentage of appeals that favor the asylum seeker. As noted at the outset, about half of all applicants eventually seek review with the BIA, but only about one-fourth of those petitions that are reviewed are overturned. The peak year for asylum appeals was 2003, with just over twenty thousand cases appealed, although, as an overall percentage of decided cases, the rate is much higher before 1996 than after. Similarly, in general it appears that the BIA was friendlier to applicants in the early 1990s than at any other period in our data. We explore these trends more in Chapter 5. Board decisions may be referred to the attorney general for review at either the attorney general’s request or the request of DHS. The attorney general may vacate any BIA decision and instead issue his or her own decision. Applicants for relief may seek judicial review of final agency decisions with the appropriate federal circuit court. In the past decade, it has been the voice of federal circuit judges ( joined by the media and the law profession) that has raised serious questions about the asylum process, specifically in relation to IJs.

14

Chapter 1

Table 1.2. Board of Immigration Appeals, Caseload and Outcomes

Year

Number of Asylum Cases Appealed

Percentage of IJ Asylum Decisions Appealed

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

7,025 4,596 4,615 5,116 6,806 11,193 12,622 10,725 11,053 13,490 12,966 13,085 16,899 20,067 17,562 16,092 13,760 12,241 10,833 9,212 7,057

58 55 52 58 57 50 41 38 39 49 49 49 51 52 50 50 43 43 43 41 35

Percentage of Decisions Pro–Asylum Seeker 49 23 31 35 36 28 28 34 36 34 31 25 26 35 26 26 24 25 27 33 27

Immigration Judges: Controversies and Issues Law professors and the media have increasingly scrutinized immigration courts and the behavior of IJs. This attention has been fueled by “federal appeals courts around the country complain[ing] of a pattern of biased and incoherent decisions on asylum and rebuk[ing] some immigration judges by name for ‘bullying’ and ‘brow-beating’ people seeking refuge from persecution” (Bernstein 2006; see also Legomsky 2007).8 In March 2007 the New York Times reported confirmation from the EOIR that “11 of the nation’s roughly 215 immigration judges had been temporarily suspended from courtroom duties since June, ‘based on concerns about how they were conducting immigration proceedings’ ” (Bernstein 2007). Bernstein’s story of New York IJ Jeffrey Chase illustrates some of the frustrations experienced by IJs as well as the

Introduction

15

attorney general and the federal circuit courts. Judge Chase, who was a former human rights advocate and chairman of the American Immigration Lawyers Association’s National Asylum Reform Task Force, had “rallied on behalf of people from China seeking asylum,” but as Bernstein notes “before long, incredulous tirades became his trademark in many Chinese asylum cases, according to court records and interviews with a dozen lawyers” as his frustration grew over “a pattern of boilerplate claims that he suspected had been concocted by smugglers.” The Second Circuit Court of Appeals rebuked Judge Chase in multiple decision for “ ‘pervasive bias and hostility,’ ‘combative and insulting language,’ and remarks ‘implying that any asylum claim based on China’s coercive family planning policies would be presumed incredible’ and criticized the judge’s decisions for ‘a plethora of errors and omissions’ ” (Bernstein 2007, n.p.). Judge Posner of the Seventh Circuit Court of Appeals has been a frequent critic of IJs. TRAC notes in particular Posner’s decision to vacate a particular IJ decision “based on six ‘disturbing features’ ” in which he expressed the view that “these disturbing features bulk large in the immigration cases” (TRAC 2006, n.p.). While Judge Posner cautioned that the cases before him may not be completely representative, TRAC concluded differently: TRAC’s analysis of the decisions of most of the nation’s immigration judges about tens of thousands of different asylum cases, however, provides powerful evidence that the problems of the immigration court go far beyond the failings of a few rotten apples—the individual judges criticized by Attorney General Gonzales. Rather, the examination of the case-by-case records appear[s] to document a far broader problem: a long-standing, widespread and systematic weaknesses in both the operation and management of this court. (n.p.) The criticism of IJs has been compounded by reports of significant variation in grant rates across judges, even among those serving on the same courts (Legomsky 2007; Ramji-Nogales, Schoenholtz, and Schrag 2007, 2009). For example, the U.S. Government Accountability Office (GAO) noted an example in 2008 in which the likelihood of receiving a grant of asylum from the IJ most likely to grant asylum was 420 times greater than the likelihood of receiving asylum from the IJ least likely to grant asylum in the same court (GAO 2008, 34). The U.S. Commission on International Religious Freedom

16

Chapter 1

(2008, 115) concluded that outcomes of individual asylum claims have come “to depend largely on chance; namely, the IJ who happens to be assigned to hear the case.” These disparities across courts and across judges have raised significant questions about the quality and consistency of justice in immigration courts, and for many observers and legal practitioners they present a frustrating dilemma. On the other hand, IJs and their national association have sought to draw attention to the serious institutional constraints under which they work. IJs face an unrelentingly harsh workload. The chief judge of the Second Circuit reported that IJs must finish at least five cases per business day to stay current. TRAC reports that IJs typically handle sixty-nine cases a week and must dispose of twenty-seven cases per week (TRAC 2011). And despite recent reforms, IJs face a growing case backlog with a new all-time high backlog of 267,752 at the end of December 2010 (TRAC 2011). IJs complain about the “the constant drumbeat of case completion goals,” the requirement that judges must “rule promptly at the end of the hearing in the form of lengthy, detailed and extemporaneous oral opinion with little or no time to reflect or to deliberate,” that making credibility determinations “is extremely, extremely difficult,” and, most important, that “there is not enough time to do research and adequately read about country conditions” (Lustig et al. 2008, 65– 66). In addition, IJs typically have little staff assistance; most courts are not assisted by a clerk or bailiff, and the judges often have to operate their own tape machines. Clearly, these are overburdened and underresourced courts with high stakes for applicants. The president of the judges association recently complained that “for some people, these are the equivalent of death penalty cases, and we are conducting these cases in a traffic court setting” (as cited in Becker and Cabrera 2009, n.p.). In subsequent chapters we offer an explanation for how these circumstances influence the way IJs make decisions and in part shape the disparities in granting asylum for which the IJs are often criticized. This work builds upon a nascent body of empirical literature that we address next.

Empirical Research on U.S. Asylum Outcomes Until recently, the empirical literature on U.S. asylum has largely been oriented toward international relations and thus has consistently focused on the relationship between the United States and asylum-sending states rather

Introduction

17

than the individual decision makers. The theoretical perspectives with which asylum outcomes have been approached in most of these studies have assumed that the state is a unitary actor, and thus the empirical analysis has focused on aggregate levels of asylum granted to persons fleeing major asylum-sending states. The early asylum literature in political science demonstrated the dominant role of geopolitical and material interests over human rights in U.S. asylum policy (Gibney 1988; Gibney and Stohl 1988; Gibney, Dalton, and Vockell 1992; Loescher and Scanlan 1998). Subsequent studies that have examined the effects of social and economic conditions of applicants’ countries of origins have generally concluded that U.S. interests influence grants more than human rights conditions (Hassan 2000; Gibney 2004a) and that applicants perceived to be economic migrants are more likely to be rejected.9 However, recent empirical studies paint a more nuanced picture, suggesting that while national interests may trump human rights concerns, both national interests and normative considerations influence these outcomes (Rosenblum and Salehyan 2004; Salehyan and Rosenblum 2008; Keith and Holmes 2009; Rottman, Fariss, and Poe 2009; Holmes and Keith 2010) in addition to domestic politics (Salehyan and Rosenblum 2008). Most researchers working on asylum outcomes in Western European states have also resorted to aggregate grant rates (for example, Holzer and Schneider 2001; Neumayer 2005; Hamlin 2012), but Holzer, Schneider, and Widmer’s (2000) study of asylum applications in Swiss cantons is a notable exception. In our previous work, we focused on a more appropriate level of analysis, the individual IJs’ decisions (Keith and Holmes 2009; Holmes and Keith 2010; Keith, Holmes, and Miller 2013; Miller, Keith, and Holmes 2013). This move to the level of the individual decision maker is an important theoretical break with unitary actor assumptions in the international relations (IR) literature. Similarly, Ramji-Nogales, Schoenholtz, and Schrag (2007 and 2009) examined IJ decisions, though they studied individual AO and IJ aggregate grant rates rather than the actors’ par ticu lar case decisions. Congressionally mandated data restrictions have made it impossible for nongovernmental bodies to have access to individual case factors such as applicant characteristics and types and level of evidence presented. In a rare field study of an NGO’s asylum cases, we (Keith and Holmes 2009) found that none of a variety of evidentiary factors mattered in the judges’ individual decisions once conditions that signaled U.S. interests were controlled. We are aware of only one other study that was able to gain access to such data.

18

Chapter 1

Koehn (1991) accessed fift y-nine individual files of Ethiopian asylum seekers in the New York and Washington, D.C., courts. And in our 2010 (Holmes and Keith) study of individual IJ decisions, we found that September 11 shifted the way in which judges assessed certain factors such as whether the applicant spoke Arabic or was from a state either that sponsored terrorists or in which al-Qaeda was present. While these studies have engaged the more appropriate individual level of analysis, they, like the rest of the asylum literature, do not offer a theoretical understanding of how the individual judges make these decisions, a crucial thread that we began to explore (Keith, Holmes, and Miller 2013). In that article we presented a variation of the attitudinal model of judicial behavior that we modified by incorporating a cognitive model of decision making. There we argued that some pieces of information before IJs in asylum cases are treated more objectively while others are treated more subjectively. This theoretical model allowed us to account for informational cues that influence IJ decisions while at the same time assessing the impact of national interests and human rights conditions. We continue to expand upon this theoretical approach and empirical exploration in subsequent chapters. Finally, most of the literature has conceptualized the dependent variable as a dichotomous decision, which our subsequent work (Miller, Keith, and Holmes 2013) began to question, given the more nuanced nature of U.S. asylum law, which presents the IJs with different standards of evidence depending upon the form of relief requested and different levels of benefits to the applicant depending upon the form of the relief granted. We explore this issue further in Chapter 2. Our book fi lls several significant gaps in this empirical literature and makes a significant contribution to the broader literature dealing with U.S. asylum law and policy. In the next section we discuss the theoretical perspectives of each component of our analyses. We also give a brief overview of our findings and their implications.

Our Contribution to the Literature First, we created a comprehensive new dataset on IJ decision making that spans two decades and contains the most complete set of covariates of which we are aware. In creating the dataset we addressed core limitations in data examined in most of the empirical literature on U.S. asylum decisions. For example, for a variety of reasons most of the asylum studies in the IR litera-

Introduction

19

ture have focused on aggregate grant rates. And even though empirical studies within the law literature have focused more appropriately on the individual asylum case and IJ decisions, this literature has tended to severely truncate the population of IJ asylum decisions, in terms of temporal dimension, the selection of high volume immigration courts, or a focus on select countries of origins or only affirmative applications. We created a more complete dataset, which we present in Chapter 2 and that examines all asylum cases decided on the merits between 1990 and 2010. We also created or gathered a significant set of case characteristics that allow us to leverage more information in multivariate analyses. Thus, we are able to specify our models to test assumptions about the countries of origins, the immigration courts, and various factors relating to the applicant. Another significant innovation in our dataset is that we move beyond the traditional treatment of the asylum outcome as a dichotomous choice in which the applicant is granted relief or not (see, e.g., Ramji-Nogales, Schoenholtz, and Schrag 2007; Rosenblum and Salehyan 2004; Salehyan and Rosenblum 2008; Rottman, Fariss, and Poe 2009; Holmes and Keith 2010; Keith, Holmes, and Miller 2013). While we acknowledge that this approach to the asylum process is reasonable, here we explore the potential loss of information that is theoretically and statistically important. We demonstrate in Chapter 2 that asylum outcomes are sometimes more appropriately operationalized as a polychotomous choice that reflects an ordering of benefits that adhere to the applicants, as opposed to a simple grant or deny decision or a legal ordering of the standards applied in the various forms of relief. One of our most significant innovations vis-à-vis the asylum decision making literature is to create a measure of the ideology for IJs. To create our measure of policy predispositions, in Chapter 2 we create a factor score that summarizes the contribution of a number of background characteristics to the policy predispositions of a judge toward asylum cases. This is a more policy-specific method of measuring a judge’s policy predisposition—our measure is specific to asylum decisions, and we do not intend it as a general measure of judicial ideology (although we occasionally refer to it as a measure of ideology for convenience). This variable becomes the core component of the cognitive model we present in this book. In addition to gathering a wealth of information on the factors that affect the decision making of IJs, we also collected data on which cases were appealed to the BIA as well as data on many dimensions of the local and national economic and political environments, which are theoretically important in understanding individual

20

Chapter 1

asylum decisions and U.S. asylum policy. In Chapter 2 we give an in-depth description of the data and the processes we employ to explore and create the new measures such as the ordered dependent variable and IJ ideology score, and we present an initial descriptive picture of the data and salient trends in asylum outcomes. Second, we draw upon two disparate disciplines and their literatures to present a model of cognitive decision making for IJs. The law literature has been primarily driven by the assumption that international and U.S. asylum law will be applied consistently and fairly. This literature has largely responded normatively to the evidence of significant disparity and inconsistency in grant rates within the asylum system. The IR literature, with its focus on the relationship between states, has largely been concerned with the foreign policy dimension of U.S. asylum decisions, along with the normative and behavioral questions that are raised by connection. These two literatures have largely operated in isolation from each other, and at the same time both literatures have failed to draw upon a third body of work, the judicial behavior literature, that greatly enhances understanding of asylum outcomes. In Chapter 3 we integrate these approaches to leverage their insights and supplement them with a variation of the attitudinal model from the judicial behavior literature that draws upon insights from psychological models of decision making. We believe that a comprehensive model of IJ decision making approximates a situation in which an IJ, under tremendous time pressure and unsure of the credibility of an asylum seeker, will use policy predispositions to help process both legally relevant and legally irrelevant facts. By incorporating a cognitive model of decision making, we show how IJs consider some pieces of information objectively while other information is treated subjectively. This approach allows us to account for informational cues that influence decisions while assessing the impact of national interests and human rights interests as reflected in the IR debate. In brief, we draw upon Braman and Nelson (2007), among others (e.g., Simon 2004; Bartels 2010), and argue that the inherent lack of certainty in legal standards, the frequent lack of corroborating evidence, and the complexity of credibility determinations invite the use of motivated reasoning in which policymotivated goals overwhelm accuracy goals. More specifically, directional or policy goals will induce what Bartels (2010) has termed top-down decision making in which the judge brings a theory or predisposition to the case. In contrast, in a bottom-up approach facts are evaluated deliberately and deci-

Introduction

21

sion making is constrained by the manner in which the law dictates the treatment of those facts. In essence, we posit that the policy preferences of IJs influence their decisions in asylum cases, but that U.S. asylum law also imposes some constraints on the use of the policy proclivities of an IJ. We explore the theoretical links fully in Chapter 3. We find that the influence of these competing factors is conditional upon a judge’s policy preferences toward asylum. More specifically, we find that the policy predispositions of the IJs play a dominant role in explaining the discrepancies in asylum grant rates that have disturbed law professors. Moreover we find that liberal IJs respond to certain nonlegal factors differently than their more conservative colleagues—findings that strongly implicate norms of fairness and consistency. At the same time, we also find some limited evidence that the law constrains the decision making of the IJs with respect to applicant characteristics. The analyses in Chapter 3 have theoretical implications and implicate several salient policy issues. In terms of theory, the analyses expand and improve upon the compliance literature by examining actual on-the-ground state actors, who engage in behavior controlled by U.S. international commitments. While previous studies in the IR literature have shown the role of both national interests and humanitarian concerns, our cognitive model offers explanations of how these factors play out in the IJs’ decision-making process. Thus, the underlying implication for political science is the importance of cross-discipline research, in the sense that IR theories benefit from the introduction of judicial behavior theories by specifying an underlying causal mechanism for many of the central IR fi ndings in what is perceived as “state” action. In regard to the work of law professors, the research in Chapter 3 strongly emphasizes the need for theory to understand data. To our knowledge, this is the first time the variation in grant rates among IJs has been examined using theoretical approaches from the study of judicial behavior. We also make two contributions to the judicial behavior literature. First, we offer an extensive exploration of the application of the attitudinal approach in a trial-level-like context where policy preferences are not thought to explain a good deal of variation in judicial decision making. Here we find that attitudes can play a prominent role in decision making given the right context (see Zorn and Bowie 2010). Second, we offer an application of psychological approaches to understanding judicial choices. Our cognitive model is innovative because it tells us not only that policy preferences matter, but also how and when they matter. In this way we also provide some evidence for the constraint of law in judicial decision

22

Chapter 1

making, which some scholars characterize as a central question in the literature (e.g., Braman 2010; Segal and Spaeth 2002). Within the asylum literature there is no work that we are aware of that examines whether local economic or demographic conditions have an impact on asylum decisions, unlike the broader immigration literature’s competing threat and contact theories that are very much focused on the local context. In this literature the primary debate centers on whether increased interaction with immigrants can either incite hostility or promote acceptance. We expand our cognitive model in Chapter 4 to integrate expectations derived from long-standing debates within the immigration literature to further understand the asylum decisions of IJs. We argue that understanding local influences on IJ decision making is important if we are interested in understanding and reducing the variability between adjudicators and across immigration courts. In the context of IJ decision making a very particular set of local conditions is likely to influence the IJ’s decisions, and these extralegal factors, similar to those related to material and security concerns we examine in Chapter 3, are likely to be evaluated in a more topdown fashion or more subjectively than legal factors and thus will be contingent upon the IJ’s policy predispositions. In other words, we believe that IJs might weigh these locally situated extralegal factors differently, depending on their own ideological proclivities. We find evidence that IJs are responsive to local conditions in highly conditional ways in that liberal and conservative IJs react differently to the same set of local conditions. This finding has important implications as not just locality influences the chances of an asylum seeker being granted asylum, but also within a community the random draw of a judge influences how those local conditions affect the applicant’s chances of relief. If IJs not only react to different local conditions but also react differently to the same local conditions, then significantly reducing the variability in their decision making may be more difficult than first contemplated. We also broaden our focus outside the immigration court to the BIA, which serves as the highest administrative tribunal adjudicating immigration matters. For many asylum seekers, the BIA has historically been their last and best chance to challenge a final deportation order. For IJs, the BIA is the most immediate venue of review in which their decisions have historically faced a real possibility of being overturned or remanded; however, due to significant reforms in the DOJ, IJs are now considerably less likely to have their decisions disturbed than they were previously. Our contribution here

Introduction

23

is primarily in examining how the behavior of the BIA might alter the decision making of the IJs and asylum applicants. In Chapter 5 we first examine the institutional context in which the BIA operates, particularly examining changes made by recent attorneys general in regard to the BIA’s structure and its methods of decision making. These reforms have been criticized within the legal literature and by practitioners and adjudicators at all levels including the federal circuit courts. In Chapter 5 we present a theoretical model for the asylum seekers’ decision to appeal to the BIA drawing upon the rational actor approach, and we find strong evidence of rationality on the part of asylum applicants. In Chapter 5 we also examine BIA outcomes themselves in what we believe is one of the more rigorous statistical examinations in the literature. We approach BIA decisions from a functional perspective, as our interest in this book focuses primarily upon the applicant and the IJ. Thus we are more interested in how the BIA informs our understanding of the individual asylum seeker’s behavior and more important how it may shape IJ behavior. Drawing upon the judicial behavior literature, we examine the BIA decisions from an error correction perspective as well as from a legal policy making perspective. While we find support for both functions, we find stronger support for the policy making function of the BIA over its error correction function. Our findings in regard to this executive branch appellate institution are consistent with the judicial behavior literature. Another contribution we make in Chapter 5 is to enhance our understanding of the role of the attorney general in the asylum process through the creation of regulations concerning the structure of the BIA and its procedures. We demonstrate that Ashcroft’s streamlining reforms clearly changed the outcomes of the BIA and had implications for both the policy-making function as well as the error correction function of the BIA. We also find evidence that overall the Ashcroft streamlining reforms appear to have polarized the BIA ideologically. We then are able to apply this understanding to our cognitive model of IJ decision making. Here we find some evidence that the IJs behave strategically in response to institutional changes put into place by Ashcroft and Gonzales, but once the streamlining reforms are instituted and affirmance becomes the norm with the BIA, IJs resume voting their policy preferences. One of our biggest contributions to the asylum literature is to broaden the perspective beyond that of the asylum applicant, in which we focus on the probability of individual relief, to that of the policy maker and system

24

Chapter 1

outputs. In Chapter 6 we examine key statutes passed by Congress in the past two decades in regard to asylum. Following upon the 1993 World Trade Center bombing, Congress passed IIRIRA, which substantially overhauled the asylum process. Then following the September 11 attacks Congress passed the Real ID Act, which sought to put tighter restrictions on potential terrorist migrants and those coming to the United States for economic, as opposed to humanitarian, reasons. We turn to the broader immigration literature and examine the laws’ intent and consequences from the policy gap perspective. This approach highlights “significant and persistent gaps [that] exist between official immigration policies and actual policy outcomes” (Cornelius and Tsuda 2004, 4) that stem from the embedded economic realities of U.S. labor market needs or from principal-agent problems in regard to implementation. In Chapter 6, to test the policy gap perspective, we identify the intent of the two statutes by examining congressional debate and public statements. We find that the rhetoric largely links the statutes to the threat of terrorism and to reducing potential abuse of the asylum system by economic migrants. We then examine our expectations in a monthly time series models, focusing on whether the intended consequences materialize, controlling for the potentially intervening shocks of the World Trade Center bombing and the September 11 attacks, along with key factors that we know from the previous chapters are likely to influence asylum outcomes. We also examine the human rights critics’ concerns that the effects of the laws are draconian and have made it difficult for bona fide asylum applicants to avail themselves of the protection promised under U.S. statutory and treaty commitments. In Chapter 6 we find overwhelming evidence to support the policy gap expectation of unintended consequences. Specifically we find that IIRIRA and Real ID had net positive effects on the number of applicants granted relief, not the negative effect that critics had feared.10 Our findings suggest that the two laws likely increased the quality of applications reaching IJs. In Chapter 7 we summarize our results and talk about what they suggest is likely to work, and how, when it comes to reforming the asylum decisionmaking process in the United States. Our focus in this final chapter is on applying our empirical insights to determine where change is likely both to be feasible and to have the intended consequences. Our conclusion there is that, generally speaking, eradicating or significantly reducing variation in outcomes for asylum applicants is likely to be difficult. We instead seek to focus attention not on the outcome but on the process of decision making. We do so by emphasizing that the quality of the determination made by an

Introduction

25

IJ should be of more concern than the actual outcome since it is likely impossible to know whether any particular applicant should be judged credible or incredible. Of course, assessing the quality of any given decision is also no easy feat (Mitchell 2010; Wistrich 2010) and considerable work, both theoretical and empirical, remains to determine how we might assess quality from outside the actual judging process. We offer some tentative suggestions for reforms that are likely to improve the quality of the decisions made by IJs, with perhaps the central recommendation being that all asylum determinations on the merits should include some brief written opinion.

CHAPTER 2

Creating a Dataset

In this chapter we discuss the creation of our comprehensive dataset, which we view as part of our contribution to understanding asylum decision making. Our dataset is focused primarily on immigration judges (IJs) because they are arguably the most important decision makers in the asylum process. We provide a comprehensive description to illustrate the significance of our improved dataset and measurements, while at the same time providing a descriptive overview of core components of the asylum decision-making process that sets up the analyses to follow in subsequent chapters. We frame the discussion in the context of the extant literature upon which we are building, identifying that literature’s specific limitations and setting out our attempts to move beyond these limitations and to create a more robust dataset.

The Population of Immigration Judge Decisions There are two major limitations in regard to the population of decisions examined in the existing empirical literature on U.S. asylum decisions: (1) some studies make assumptions about the decision-making process that may be untenable and (2) most studies have examined a nonrepresentative sample of decisions. In the international relations literature, three factors have led most of the studies to focus on aggregate grant rates for high asylumproducing countries rather than examining the individual case as the level of analysis: (1) the unitary actor assumption, (2) a focus on interstate relations, and (3) data availability (Rosenblum and Salehyan 2004; Salehyan and Rosenblum 2008; and Rottman, Fariss, and Poe 2009; but see Holmes and Keith 2010; Keith, Holmes, and Miller 2013; and Miller, Keith, and Holmes 2014).

Creating a Dataset

27

With respect to data availability, the largest available dataset for scholarly research is that created by Ramji-Nogales, Schoenholtz, and Schrag for their 2007 law review article. While these authors focus more appropriately on the individual asylum case and IJ decisions, they severely truncate the population of IJ asylum decisions. First, they focus on a somewhat narrow period: 2000 to 2004. Second, they include only cases heard in “high volume courts”1 and only cases of individuals from asylum-producing countries (APCs).2 Third, they drop all defensive cases in order to exclude instances in which the applicant was detained (Ramji-Nogales, Schoenholtz, and Schrag 2007, 395). However, not every defensive case involves a detained asylum seeker. Our correlation of defensive claims and detained applicants suggests that the measure is in fact not a strong proxy for detention status—the correlation is only moderately strong (r = .62). After excluding these three categories, their dataset contains only 66,443 cases, compared to 159,110 in our dataset when we select the same period. Their population of cases is further reduced in that only APC applicants in these high volume courts are included when an IJ decided at least one hundred cases from an APC. Thus, the sample population is restricted through a variety steps that potentially bias the dataset in favor of a grant (Ramji-Nogales, Schoenholtz, and Schrag 2007, 397). While we understand the possible benefit of such restrictions when conducting bivariate analyses, the reduced sample makes it difficult to draw generalizable conclusions from the data. In our dataset, instead of excluding cases, we examine all asylum cases decided on the merits between 1990 and 2010. The data were attained through numerous Freedom of Information Act (FOIA) requests to the U.S. Citizenship and Immigration Ser vices (USCIS) and the Executive Office for Immigration Review (EOIR). The inclusion (and control of case characteristics) allows us to leverage more information in multivariate analyses. We are able to specify our models to test assumptions about the countries of origins, the immigration courts, and various factors relating to the applicant, including detention status and whether the application is defensive or affirmative. Our decision to model each of these factors simultaneously is important because it means that each competes to explain variation in the decision making of IJs in a manner that is impossible with bivariate approaches. In all, we have 589,629 observations from 1990 through 2010, amounting to an average of 28,078 cases a year. The number of claims varies somewhat over time, a point we illustrated in the introduction. Here we wish to dig deeper into the data to give a more in-depth picture of the kinds of information we have created.

28

Chapter 2

An important variable in our analysis is the country of origin from which an alien is fleeing. There are applicants from 173 countries in our dataset, although 5 countries provide over 45 percent of the cases: China (19 percent), Haiti (8.32 percent), Guatemala (6.49 percent), El Salvador (6.47 percent), and Colombia (5.52 percent). Approximately 28 percent of the applicants speak Spanish, 11 percent speak English, and 2 percent speak Arabic. Most of the applicants come from underdeveloped countries. Using the World Bank income classification scheme, 89 percent of those applying for asylum between 1990 and 2010 come from countries classified as low income or low middle income—meaning that per capita income is less than $4,035 in 2012 dollars. This suggests that concerns over abuse of the asylum system by economic migrants may need to be at least considered, a task we take up in subsequent analyses. It is also the case, however, that most of those seeking asylum in the United States are fleeing truly repressive countries. Using U.S. Department of State reports on country human rights conditions, 52 percent of the applicants come from countries that score a 4 or 5 on the personal integrity abuse scale developed by Mark Gibney and updated with his coauthors (Wood and Gibney 2010; Gibney, Cornett, and Wood 2010).3 Essentially, this means that at least half of those seeking asylum in the United States come from countries where severe repression or egregious violations of human rights are a regular part of life and where murder and torture are common. For example, both Afghanistan and Colombia received scores of 5 on this scale in 2001 because severe levels of repression affect the entire population, not just leaders or activists. China received a 4 because severe repression, although widespread, affected mostly those who participated politically or were identified with certain ideas or goals. In contrast, a country like Cuba received a score of 3, due to common persecution of political opponents but the absence of common violations of political or civil rights among the general population. Scholars of the U.S. asylum process divide asylum claims into two primary types: affirmative and defensive. Affirmative claims are claims for asylum made by applicants before any type of action has been taken against them by the U.S. government; defensive claims are those made by aliens whom the U.S. government is seeking to remove from the United States. It is generally believed, and sometimes shown, that affi rmative applications (which are initially heard by an asylum officer [AO]) are treated more favorably than are defensive claims. We note first that the majority of claims in our data are affirmative (62 percent) and that those pursuing affirmative claims

Creating a Dataset

29

are significantly more likely to receive some form of relief than are applicants in the defensive process. Of applicants who pursue asylum affirmatively, 41 percent receive some type of relief, while just 32 percent of those who pursue it defensively receive some type of relief.4 Another factor thought to affect the ability of aliens to pursue an asylum claim is whether or not they are detained by the United States in a detention center while awaiting adjudication, as this clearly makes gathering and authenticating evidence difficult, particularly when many of those who are detained do not have legal counsel. The U.S. government data code applicants into one of three categories: never detained, currently detained, and released. For the individuals on whom we have reported data, the modal category is never detained, with 77 percent of applicants avoiding detention. In addition, 14 percent of applicants fall into the released category, meaning they were previously detained but were not detained at the time the IJ heard their claim. Finally, the remaining 9 percent of applicants were detained at the time of their hearing before an IJ. As expected, those who are in detention when their claims are heard are significantly less likely to be granted any form of relief—an applicant who is not currently detained or has never been detained is 68 percent more likely to avoid removal from the United States. The difference between those who were never detained and those who were detained but then released is also substantial: those who were previously detained but were released at the time of their hearing were 25 percent less likely to receive asylum than those who had never been detained. Also in the vein of factors that affect the ability of an alien to claim asylum, IIRIRA put in place a legal bar to the seeking of asylum: if an alien has been in the country for more than one year prior to filing for asylum, then he or she cannot be granted asylum, subject to certain exceptions.5 Using data provided to us by the USCIS, we were able to determine the time from arrival to the initial filing of an application and therefore whether an applicant was subject to this bar on asylum. The one-year deadline was enforced beginning April 1, 1997. Not unexpectedly, the number of applicants reaching IJs with asylum claims made after one year in the country decreased significantly: before the implementation of the bar 79 percent of claims were filed after one year had elapsed from arrival, whereas after the enactment of the IIRIRA restriction, 63 percent of applicants made claims that, based solely on the timing of the filing of their application, may have been subject to the one-year bar. It appears then that IIRIRA may be operating as a substantial impediment to asylum claimants, but the effect may not be as drastic as first contemplated.

30

Chapter 2

First, it is clear that applicants subject to the bar are less likely to prevail in their asylum claims by about 8 percentage points. Of applicants subject to the bar, 35 percent receive asylum, while 43 percent of those not subject to the bar receive asylum.6 Second, those proceeding subject to the ban are just over three times as likely to receive a grant of withholding of removal in lieu of asylum. When asylum and withholding of removal are added together as forms of relief, the differences between those subject to the one-year bar and those not subject to it compress somewhat: 44 percent of applicants not subject to the bar receive one of these two forms of relief, while 41 percent of those subject to the bar receive one of these two kinds of relief. Though the absolute numbers are small, it is also worth noting that 75 percent of those who receive withholding under the Convention Against Torture (CAT) are also subject to the one-year bar on asylum claims. Obviously the one-year bar is consequential for applicants because, as we will explore more fully below, they are more likely to receive a less substantively beneficial form of relief than are applicants who file within one year of arrival. Yet it also important to recognize that, even subject to this bar, many applicants, over one-third in our data, receive asylum. One caveat must be noted here—we are missing data on either the date an applicant arrived or the date his or her first application was filed for 11 percent of the post-IIRIRA sample, and because we do not know why EOIR omitted those data, any conclusions about the effects of the one-year bar must remain somewhat tentative.7

Operationalization of Dependent Variable All prior empirical studies of asylum decision making of which we are aware treat the outcome as dichotomous: either the applicant is granted relief or he or she is not (see, e.g., Ramji-Nogales, Schoenholtz, and Schrag 2007; Rosenblum and Salehyan 2004; Salehyan and Rosenblum 2008; Rottman, Fariss, and Poe 2009; Keith and Holmes 2010; Keith, Holmes, and Miller 2013). This approach to the asylum process is reasonable, but it collapses what is an ordinal variable into two categories, causing a potential loss of valuable information (Kennedy 2008, 245). In this chapter, instead of collapsing the outcomes into a dichotomous decision of any relief versus none, we investigate, as a preliminary matter, whether the multiple outcomes in asylum cases can be ordered and, if so, what explains that ordering. As noted in the introduction, there are four potential outcomes in an asylum case: no relief, with-

Creating a Dataset

31

Table 2.1. Distribution of Outcomes, 1990–2010 Outcome

Frequency

Percentage of Total

No relief Withholding under CAT Withholding Asylum

397,219 2,049 20,741 169,620

67.37 0.35 3.52 28.77

Total

589,629

holding of removal under the CAT, withholding of removal, and asylum. Table 2.1 below presents the distribution of these outcomes in our data.8 To be sure, the categories between no relief and asylum—the two types of withholding—compose just under 4 percent of the data, and therefore one might argue that collapsing them into the relief side of the no relief/relief dichotomy does little harm.9 Whether or not it is wise to do so is an empirical question with important substantive and policy components—a question we answer below. Empirically, to determine whether the dependent variable is ordered we need to test whether such ordering is reliable in the data and on what dimensions any ordering might occur. There are two potential orderings of these data: one based on the substantive benefits awarded to an applicant and one based on the legal standards that must be met in order to grant a specific type of relief. Determining which of these orderings best describes the manner in which IJs approach their task is of substantive interest. Each form of relief is decided according to a different legal standard under U.S. law. The most significant criterion for obtaining asylum status is the basic refugee definition in the 1980 Refugee Act, which is based on the Refugee Convention Article 1 qualification for a refugee (asylee in this context): “[A]ny person who is outside any country of such person’s nationality . . . and who is unable or unwilling to return to, or is unable or unwilling to avail himself or herself of the protection of, that country because of persecution or a well-founded fear of persecution on account of race, religion, nationality, membership in a particular social group, or political opinion.” U.S. law also delineates bars to eligibility of asylum status, such as being a “persecutor of others,” being convicted of an aggravated felony, having previously fi led for and been denied asylum, or “posing a danger to the security of the United States.” Withholding of removal or non-refoulement is based on the

32

Chapter 2

Convention’s Article 33 nonreturn obligation and does not confer asylum status—rather individuals granted withholding of removal live with a final deportation order fi led by the IJ. Consequently, if the individual leaves the United States after being granted withholding of removal, this action in effect “self-enforces” that deportation order and the individual cannot ordinarily return until a three- or ten-year bar to reentry passes (Anker 2011, 8). The Refugee Act makes withholding of removal a mandatory protection. We believe the following legal ordering exists: (1) no relief, (2) asylum, (3) withholding of removal under the CAT, and (4) withholding of removal based on the standard of proof. Granting of asylum is entirely discretionary under U.S. asylum law; whereas, withholding of removal and withholding under the CAT is not and the standard for giving protection is that the odds of future persecution are “more likely than not.”10 One way to conceptualize this distinction is to note that to get asylum an applicant needs to prove a possibility of persecution; whereas, to get any type of withholding the applicant must prove a probability of persecution upon return. While the standards of proof make withholding of removal and withholding under the CAT less likely, there are legal dimensions of these protections that broaden somewhat the possibility of protection. Unlike in the case of asylum claims, neither CAT nor withholding claims are subject to the one-year time limit. To distinguish further between the two types of withholding, we note that withholding under the CAT has no bars to eligibility (thus protection can be granted to criminals, terrorists, and persecutors), and while future torture is the only form of persecution from which an applicant is protected, the torture need not be linked to one of the five protected grounds. This is not true of general withholding of removal, which does have bars to eligibility (such as applying after the one-year deadline in U.S. law), and furthermore, an applicant must prove that persecution is linked to one of the required protected grounds. In all, then, we believe that legally ordering the variables is somewhat difficult because the standards of proof, the various bars to protection, and the types of persecution required for withholding and withholding under the CAT are not easily ordered. Table 2.2 presents a summary of the legal standards as well as the substantive benefits across the three forms of relief, which we discuss next. Alternatively, the ordering may be based on the substantive benefits granted to an applicant given the awarding of a particular form of relief. In terms of relief, individuals granted asylum are permitted to remain in the United States for an indefinite period and may apply for permanent resident

Table 2.2. Legal Standards and Substantive Benefits Across the Three Forms of Relief

Asylum Legal Standards Protection Standard of proof (odds of future persecution)

One-year limitation of IIRIRA Bars to eligibility Limits on type of persecution

Entirely discretionary Reasonable possibility standard (must prove a possibility of persecution) Subject to one-year limitation Yes No

Substantive Benefits Status Permitted to remain for indefinite period Can apply for permanent resident status after one year

Family

Other repercussions

Asylum granted to family members included in application May petition to bring eligible family members to the United States None

Withholding of Removal Under the CAT

Withholding of Removal

Mandatory

Mandatory

More likely than not standard (must prove a probability of persecution)

More likely than not standard (must prove a probability of persecution)

Not subject to one-year limitation No Only torture

Not subject to one-year limitation Yes No

No status—just protection against being returned to the persecuting country No path to apply for permanent resident status Eligible to apply for work authorization No relief for family members No ability to petition to bring family over

No status—just protection against being returned to the persecuting country No path to apply for permanent resident status Eligible to apply for work authorization No relief for family members No ability to petition to bring family over

Can be detained after granting relief Can be removed to a safe third country

Face future removal if DHS reopens the case due to changing country conditions Cannot leave United States or selfenforce final deportation

34

Chapter 2

status after one year. In addition, asylum relief is also granted to the asylee’s present family members who were included in the asylum application, and asylees may petition to bring eligible family members to the United States. Withholding of removal offers no status, just protection against being returned to the persecuting country. Recipients do not gain relief for family members, the ability to bring family over, or a path to lawful permanent residence. Withholding does, however, give recipients (but not their family) the ability to apply for work authorization. Similar to withholding of removal, relief under the CAT does not provide a path to become a lawful permanent resident or relief for family members. It does allow recipients to apply for work authorization, but at the same time allows the United States to detain CAT recipients. Based on this variation in benefits, the expected ordering would be (1) no relief, (2) withholding under the CAT, (3) withholding, and (4) asylum. Which ordering makes more sense given the data? We are able to utilize stereotype logit models (Anderson 1984; Jones and Westerland 2006) to estimate, given a fully specified model, parameters (phi) that measure the distinguishability of the categories of the dependent variable (Long and Freese 2006).11 Essentially, this tests the ordinality of the dependent variable (Jones and Westerland 2006) and can help to determine if either proposed ordering is supported in our data. Put differently, instead of simply imposing an ordering on the data, the stereotype logit model allows the data to determine whether such an ordering makes sense. In addition to estimating the effects of variables on the dependent variable, the process also allows us to estimate parameters of the ordering. Therefore, we can conclude if a particular ordering is supported by the data (given the model) and whether the individual choices contained therein are distinguishable from one another. These parameters help to scale each of the choices on a common metric. What matters is the relative ordering of the phi parameters—if an ordering fits the data, then phi1 > phi2 > phi3 > phi4. This relative ordering means that the choices made by IJs accord with our a priori assumptions about the ordering. Below we estimate two separate models, one with a substantive ordering and one with a legal ordering, to test whether either set of orderings fits with how IJs tend to decide asylum cases.12 The order chosen by the data is premised on a model of the datagenerating process—and the results we present are premised on the cognitive model of judicial decision making we specify in Chapter 3. Here we present the relative orderings of the phi parameters for the fully specified cognitive model, where the substantive ordering is (1) no relief, (2) withhold-

Creating a Dataset

35

ing under the CAT, (3) withholding, and (4) asylum and the legal ordering is (1) no relief, (2) asylum, (3) withholding under the CAT, and (4) withholding. Evidence for ordinality comes from the ordering of the phi parameters, with the two “corner” phi parameters constrained to 1 and 0, respectively (this is necessary for identification of the model)13—that is, phi > phi > phi > phi. It is important to note that the stereotype logit model does not impose this ordering on the data (as in an ordered probit model), but instead allows the application of the model to the data to determine which of the potential orderings best fits the data. It is clear from the evidence in Table 2.3 that the substantive benefits ordering is supported by the data, but the legal ordering is not.14 For the substantive benefits ordering, each category is distinct (there is no overlap in the estimate 95 percent confidence intervals for the categories), and the ordering is as predicted. Furthermore, the distances between the parameters here support our relative orderings, with a grant of withholding close to a grant of asylum, but a grant of withholding under the CAT close to a decision to grant no relief at all. In the legal ordering, the phi (asylum) and phi (withholding under the CAT) parameters are jumbled and phi (asylum) is potentially distinct from the other three outcomes. In its way, then, the legal ordering supports the substantive ordering. This is important because it suggests to us that IJs are deciding cases more on the basis of the benefits available to the applicant than on the basis dictated by U.S. asylum law. This is but the first instance we will report of IJs innovating in their decision making, often in ways that disregard the law that is supposed to guide their decision making. Furthermore, we can make substantive sense of the distances between the phi parameters (Long and Freese 2006). Recall that in the benefits ordering Table 2.3. Substantive Benefits and Legal Ordering Substantive Benefits Ordering Phi Phi Phi Phi

No relief Withholding under CAT Withholding of removal Asylum

Legal Ordering

1 0.75 [0.58, 0.91]

No relief Asylum

1 0.07 [−0.06, 0.19]

0.18 [0.06, 0.30]

Withholding under CAT Withholding of removal

0.41 [0.27, 0.55]

0

0

36

Chapter 2

phi represents a decision not to grant any form of relief and that phi represents a grant of asylum. Withholding under the CAT appears to be treated as substantively more similar to a denial of any form of relief than to a granting of asylum, given that the parameter estimate for phi (which represents withholding under the CAT) is 0.75 (which is closer to the value of the phi parameter than to the phi parameter). In contrast, withholding of removal (not under the CAT), here represented by phi, is 0.18, which is quite close to a grant of asylum. The results suggest that withholding under the CAT is considerably different from withholding and that withholding is a more substantively beneficial result than is withholding under the CAT, at least in the minds of IJs. Finally, it is worth treating the dependent variable in the analyses that follow as ordered because variations in the use of the various categories of relief over time tell us important information about the strategies used by U.S. policy makers to limit or discourage the pursuit of asylum by migrants. As we noted in the introduction, the use of the various categories of relief changes over time, so that, although withholding of removal only occurs about 4 percent of the time in all of the data, by the latter part of the 2000 to 2010 decade it constitutes almost 8 percent of all decisions. Therefore, the finding that it fits somewhere between a denial of all relief and asylum on a substantively ordered scale is telling because it suggests one method by which the IJs have continued to allow asylum seekers some level of protection while simultaneously denying applicants many of the substantive benefits available to asylum recipients. This trend fits well with the findings of other scholars who have studied the response of Western democracies to increasing pressures on their immigration systems, and specifically their asylum systems (Böcker and Havinga 1998; Schuster 2000; Gibney 2004a, 2004b).

Measuring the Policy Proclivities of Immigration Judges One of our major innovations vis-à-vis the asylum decision making literature is to create a measure of the ideology for IJs. One typical approach to measuring judicial ideology is to use the ideology or party of the appointing president as a proxy for the ideology of the judge, in various guises (see, e.g., George and Epstein 1992; Segal, Howard, and Hutz 1996; Zupan 1992), including using these proxies specifically in asylum cases (Ocepek and Fetzer 2010). To create our measure of policy predispositions, we generate a factor

Creating a Dataset

37

score that summarizes the contribution of a number of background characteristics to the policy predispositions of a judge toward asylum cases. This is a more policy-specific method of measuring a judge’s policy predisposition— our measure is specific to asylum decisions, and we do not intend it as a general measure of judicial ideology (see Fischman and Law 2009 for a further discussion of this distinction). A subject-specific approach such as ours is often necessary when scholars study specialized judicial decision makers because more general approaches may not uncover the relevant ideological dimensions of decision making (Baum 1994; Staudt, Epstein, and Wiedenbeck 2006; Miller and Curry 2009, 2013). Our approach builds upon earlier studies that have looked to judges’ background characteristics, believing that they represent a socialization process that ultimately results in votes (Gryski, Main, and Dixon 1986). The prior experiences of judges serve as “indicators or clues about life experiences and judges perspectives toward cases” (Tate and Handberg 1991, 460). In addition, there is strong evidence that the background characteristics that we consider have been used by key actors in the asylum decision making bureaucracy specifically as cues for the likely policy preferences of asylum adjudicators. For instance, when Janet Reno sought to increase the size of the Board of Immigration Appeals (BIA) in the 1990s, she was careful to select members who had backgrounds in academia, private practice, and advocacy on behalf of immigrants to balance what was thought to be an overwhelming dominance by those with a background in government enforcement of immigration laws (see Schoenholtz 2004–5). Furthermore, when Ashcroft sought to streamline the BIA he removed those judges who had previous experience working in NGOs and on behalf of immigrants. To this end, we examine the IJ’s career path to create a tightly focused proxy for a policy predisposition toward immigration rights and asylum. We believe that our measure is a strong proxy for asylum liberalism that likely reflects a socialization process that we discuss further below. In addition, we believe that it accounts for the early career selections of some IJs that may indicate an underlying policy proclivity that is subsequently strengthened through additional career socialization. For example, a conservative individual may be more likely to seek out a job as an Immigration and Naturalization Ser vice (INS) agent or a prosecutor. These career experiences then are likely to reinforce those underlying proclivities. First, using unrotated factor analysis and several background variables related to the IJs, we create a factor that we believe captures an underlying

38

Chapter 2

immigration liberalism that will drive decision making. In essence this is a measure of policy predisposition, although we believe that it is most accurately described as capturing the degree to which an IJ has been socialized to give the benefit of the doubt to asylum seekers. The extent to which an IJ is willing to give asylum seekers the benefit of the doubt can be telling given the centrality of credibility determinations in IJ decision making. We included the following variables in the factor analysis: experience at INS, experience at the Department of Homeland Security (DHS) (other than INS), experience at the EOIR, experience as a prosecutor, experience at an NGO, experience at an immigration-related NGO, military experience, academic experience, private practice experience, prior judicial experience (other than as an IJ), and corporate experience.15 These variables are included because previous scholarship and personal observation of these judges lead us to believe that these characteristics are strong proxies for the liberalness of a judge. For most of the characteristics we have a priori expectations about how the characteristic will relate to asylum or immigration liberalism. For instance, we believe that experience in the DHS, at INS, or as a prosecutor should indicate skepticism toward asylum claimants. On the other hand, it is straightforward to expect that those who have worked for NGOs or for immigration NGOs more specifically are highly likely to view asylum claimants sympathetically, given that many of these NGOs represent or support the representation of immigrants. As is common practice, we retain only the factors that have eigenvalues above 1 (Gorsuch 1983). We then used these elements to create a factor score using regression techniques. The results of the factor analysis and regression scoring method are displayed in Table 2.4. The retained factor has an eigenvalue of 1.42 and explains 78 percent of the variance in the data. No other factor in the data approaches an eigenvalue of 1. The factor loadings for the various background characteristics are exactly as we would predict, with previous INS, DHS, EOIR, prosecutorial, and military experience suggestive of a conservative view toward asylum seekers. Alternatively, NGO experience, immigration-specific NGO experience, and academic experience are all predictably suggestive of increased liberalism toward asylum seekers. Prior judicial experience is slightly indicative of greater asylum liberalism, as are prior private practice experience and corporate experience. We did not have strong a priori predictions for these three variables, but their inclusion seems to explain some variance in asylum liberalism.16

Creating a Dataset

39

Table 2.4. Factor Score for Asylum Liberalism Factor Loadings

Factor 1

INS experience DHS experience EOIR experience Former prosecutor NGO experience Immigration NGO experience Military experience Academic experience Private practice experience Prior judicial experience Corporate experience

−0.414 −0.226 −0.119 −0.098 0.685 0.679 −0.178 0.331 0.298 0.019 0.092

Regression Scoring Coefficients

Factor 1

INS experience DHS experience EOIR experience Former prosecutor NGO experience Immigration NGO experience Military experience Academic experience Private practice experience Prior judicial experience Corporate experience

−0.184 −0.097 −0.034 −0.025 0.359 0.363 −0.058 0.120 0.129 0.007 0.034

As a check on the robustness of this approach to measuring the proclivities of IJ toward asylum cases, we also create a measure that we call presidential liberalism, which is a factor score combining a dummy variable for appointing president and the DW-NOMINATE score from Poole and Rosenthal (1997). Our method for creating this factor score is identical to the method described above, and we do not report the factor loading here, although we again retained one factor with an eigenvalue above 1. Our two measures of judicial policy proclivity are positively correlated, with a modest correlation of r = .15. We included this variable in alternate versions of our cognitive decision making model, as presented in Chapter 3, as a robustness check on our results. We found that the president-based scores do a much poorer job of explaining IJ decision making than do more asylum-specific

40

Chapter 2

background characteristics (see Fischman and Law 2009; Keith, Holmes, and Miller 2013). Our measure of asylum liberalism ranges from 0 to 4.08, with higher scores indicating a greater proclivity to grant relief to applicants in asylum hearings. The median score is 0.845, meaning that most of the IJs in our data lean toward conservatism in the granting of asylum. Put differently, the majority of IJs have backgrounds that suggest that they will look skeptically at asylum claims. The corps of IJs becomes slightly more liberal over time in our data, with the median asylum liberalism score increasing from 0.954 in 1990 to 1.061 in 2010. There is also some difference in the types of IJs appointed by Democratic and Republican administrations. The median score for an IJ appointed by a Democratic administration is basically indistinguishable from the median score for IJs appointed by Republican administrations. However, as the scatterplot in Figure 2.1 demonstrates, Democratic administrations are much more likely to appoint extremely liberal IJs than are Republican administrations. As further proof, 93 percent of extremely liberal IJs—those with scores above the 90th percentile— are appointed during Democratic administrations. Interestingly, however, Democratic administrations are also more likely to appoint extremely conservative IJs—those in the 10th percentile of asylum liberalism—as 73 percent of these conservative IJs are appointed by Democratic administrations. In short, for reasons that remain unexplained, Democratic administrations appear to appoint IJs with much more heterogeneous preferences than Republican administrations, although the Democratic administrations also tend to skew their appointments heavily toward more liberal IJs. Our data also allow us to speak to questions about just how secure IJs are in their tenure. IJs are civil ser vice employees in the Department of Justice, meaning they have some protections from dismissal. However, they do not, as we discuss elsewhere, enjoy the same kinds of protections as do federal Article III judges or Article I administrative law judges. Therefore, an important question is whether IJs have reason to fear dismissal for decisions that are disagreeable to their superiors. Our data indicate that an IJ’s tenure averages just under 14 years. But this estimate is likely low because our data truncate many IJ careers in 2010, and we could expect that at least some IJs would serve well beyond the end date for our data. Th is is shorter than the twenty-year average length of ser vice for federal judges (Yoon 2003), but whether the six-year difference is to be considered substantial depends on the age at which judges on each set of courts are appointed (undoubtedly higher

41

0

Asylum Liberalism (Higher is More Liberal) 1 2 3

4

Creating a Dataset

1970

1980

1990 Year IJ Appointed

2000

2010

Figure 2.1. Scatterplot of IJ Asylum Liberalism by Year Appointed.

for federal courts), the degree of truncation in our data (which is high),17 and variation in levels of career satisfaction.

Appealing the Asylum Decision In addition to gathering a wealth of information on the factors that affect the decision making of IJs, we also collected data on which cases were appealed to the BIA. Thus, we are able to see if there are systematic differences between cases that are appealed and those that are not. To begin with, we note that 47 percent of the cases in our dataset are appealed, a relatively high rate of appeal, although not entirely surprising given the incredibly high stakes for applicants in many of these cases. In these appeals, applicants are successful in 29 percent of the cases, where we defi ne success as winning some type of relief or the right to a new trial.18 IJs are reversed approximately 14 percent of the time. In addition, we note that applicants are the appealing party in 95 percent of the cases, with DHS appealing only about 3 percent of the time. The remaining 2 percent of

42

Chapter 2

appeals were initiated either by both the asylum seeker and DHS or by a third party. Although we do present a fully specified model of the decision to appeal an asylum decision in Chapter 5, we believe that a few of the correlations between case factors and the decision to appeal, and the final appellate outcome, are worth detailing here as a preliminary matter. Furthermore, these correlations are likely to have implications for various policy recommendations made for improving the U.S. asylum system. Commentators have lamented the fact that many asylum applicants proceed without representation or with poor representation (e.g., McKeown and McLeod 2009). Our data show that most asylum applicants whose cases are decided on the merits are in fact well represented—79 percent of the cases have a representative.19 Not surprisingly, applicants who are represented are much more likely to receive some type of relief, as those without representation are denied any form of relief 79 percent of the time, compared to only 58 percent of the time for those represented. Furthermore, applicants who are represented are especially likely to appeal their cases: those who are represented appeal 50 percent of the time, compared to those who are unrepresented who appeal 37 percent of the time. Asylum seekers who are represented are also significantly more likely to win their appeals. Those who are represented win 30 percent of the time, compared to 23 percent of the time for those who are unrepresented.20 What this implies is that increasing the rate at which asylum applicants are represented may have the unintended consequence of increasing the caseload faced by the BIA. If we assume that all asylum applicants were represented and that the rate of appeal remained the same as it is currently for those who are represented, then full representation would have resulted in an additional 13,800 appeals to the BIA. Whether this constitutes a significant increase in the caseload of the BIA is debatable, but we believe that consideration of this increase must be part of a discussion of the benefits and costs of increased representation for asylum seekers. Another interesting association is that between case outcomes at the IJ level and the decision to appeal. Table 2.5 presents the cross-tabulation between case outcome at the IJ level and the decision to appeal. Not surprisingly, most appeals (95 percent) are of cases in which the IJ granted no relief, yet only 5 percent of cases in which asylum is granted are appealed. This is not surprising, and the ordering presented in the table lends additional support to our assertion that there is an underlying ordering of outcomes that is premised on the substantive benefits attaching to a particular outcome, with

Creating a Dataset

43

Table 2.5. Cases Appealed, 1991–2010 IJ Outcome

No Relief

Withholding Under the CAT

Withholding

Asylum

Total

No Yes

91,411 (23%) 305,808 (77%)

940 (46%) 1,109 (54%)

16,470 (79%) 4,271 (21%)

160,744 (95%) 8,876 (5%)

269,565 (46%) 320,064 (54%)

Total

397,219

2,049

20,741

169,620

589,629

Case Appealed

withholding under the CAT appealed much more frequently than withholding. Indeed, regression models—not presented here—indicate that the outcome at the IJ level is the single most important factor in determining whether an appeal is taken, as we discuss in Chapter 5. There is a general downward trend in the number of cases appealed over time from a high of 58 percent of all cases in 1990 to a low of 35 percent of all cases in 2010, and there has been some fluctuation over time in this trend. In contrast, the percentage of cases appealed that were filed by asylum applicants has remained more or less constant, hovering between 91 percent and 98 percent of all cases fi led. The percentage of cases appealed that are won by asylum applicants has also exhibited a strong tendency to fluctuate around 30 percent, with a high for applicants of 49 percent in 1990 and a low of 23 percent in 1991. Figure 2.2 plots these three trends over time and gives a sense of the relative stability for each series, with some slight downward trends for appeal rates and win rates (which are perhaps the result of changes made to the BIA by John Ashcroft), but almost no change in who files the majority of cases. We return to these time trends in Chapters 5 and 6, wherein we also examine how the IJs have reacted to changes over time, including a series of policy changes.

Data Origins and Limitations We gathered the bulk of our data via a series of FOIA requests to the USCIS and the EOIR between 2009 and 2012. In addition, we supplemented the data with information widely available, such as the measures of personal integrity abuse from Mark Gibney, among others. We are interested only in

Chapter 2

20

40

Percentage of Cases 60 80

100

44

1990

1995

% of Cases Appealed to BIA

2000 Year

2005

2010

% of Cases Filed by Alien

% of Cases Won by Alien

Figure 2.2. Cases Appealed, Cases Filed by Alien, and Cases Won by Alien in the BIA, 1990–2010.

decisions in cases on the merits for asylum claims. We count all cases in which the alien made a claim for asylum, withholding of removal, or withholding of removal under the CAT and the IJ made a decision on the merits of that claim. This means that we do not have data on asylum claims that were dismissed for procedural reasons, such as a failure to appear. In an attempt to make inferences that are as generalizable as possible, we include all of these cases, regardless of the country of origin of the applicant or the number of cases decided by a particular IJ. One difficulty created is that our data have been gathered by two different agencies that record and report their data differently: this largely stems from the fact that each counts dependents of an applicant differently. Therefore, we are careful throughout not to directly compare the aggregate data we generate for AOs with the data for IJs; instead, where AO data are used we are interested in it primarily as a control. In general, we believe that any trends in the AO data that correlate with the IJ data should not be driven by the differences in how the two systems report data, though there will be differences reported in the relative levels of applicants.

Creating a Dataset

45

Given that we are reliant upon FOIA requests to generate our dataset, there are some inherent, government-imposed limitations on the data. Congress does not allow either USCIS or EOIR to report a host of information on the decision makers of interest, including any information that would allow us to identify the individual decision of either AOs or members of the BIA. In addition, the government does not regularly report the age or gender of applicants, although rarely this is present for some applicants. There is no information about the quantity or quality of evidence provided by the applicant. Similarly, we do not have demographic or socioeconomic information about the individual asylum seeker. The government reports the applicant’s country of origin and language spoken and whether or not he or she was represented by a lawyer. For the IJs, we have identifying information that we have coded from the biographies that the EOIR publicly releases upon their appointment. We use this to create the ideology measure that we discussed at length earlier in this chapter. Where we could not find biographical information on judges from the official EOIR published biographies, we exclude the judges. We located biographical information on as many judges as possible. First, some cases have the IJ identified as “visiting IJ.” Although there may have been numbers or cities associated with the visiting judges, we were unable to identify the specific judge. Second, of the judges who were identified by name, we were able to find biographical data on 330 of the 374 nonvisiting judges who appear in our data. We think it is important to investigate the 44 IJs for whom we were not able to locate biographical data. First, the majority of the decisions by these judges occur quite early in our dataset, with 62 percent of them happening before 1997. Second, though we are missing data on 12 percent of our IJs, the IJs for whom we are missing data did not decide many asylum cases: just 3 percent of the total observations come from these IJs. Finally, to the extent that the IJs for whom we are missing data decide cases differently than do those for whom we have data, our results need to be adjusted. We find that the IJs who have missing biographical data are somewhat more likely to decide cases conservatively than are those for whom we are able to locate data (about a 14-percentage-point difference in the likelihood of denying any form of relief ). But much of this difference stems from the fact that those judges with missing data are also significantly more likely to hear cases from those applicants who are detained at the time of their hearing (by about 7 percentage points). Given that we control for detention status in our models and limit many of our analyses to the post-1997

46

Chapter 2

period for reasons explained in subsequent chapters, and that the actual amount of missing IJ ideology data is quite small, we are not overly worried about these differences between those judges on whom we have located data and those on whom we have not. Data on this missingness, by year, is reported below. In addition, we made some assumptions on coding other variables that did not come with complete data from our sources. One of our variables measures whether an applicant was detained at any point in the asylum process. Data from EOIR provided information on 480,514 of our observations, but not on the remaining 30,624 observations for which we have IJ data. For these observations we imputed the median value of all observations. This amounts to treating these missing observations as though they were not detained. Table 2.6. Number of Decisions from IJs Without Biographical Data

Year

Number of Missing Votes

Cumulative Percentage of IJs Without Biographical Data

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

2,372 1,454 1,681 1,543 1,336 1,963 1,568 547 322 879 567 551 669 662 690 609 457 475 318 295 255

12.35 19.91 28.66 36.69 43.65 53.86 62.03 64.87 66.55 71.12 74.07 76.94 80.42 83.87 87.46 90.63 93.01 95.48 97.14 98.67 100.00

Total

19,213

Creating a Dataset

47

Although we are sure that this assumption codes some cases as not detained when in fact the applicant was detained, we tested all of our results to the robustness of this assumption, and everything we report is robust to this our coding decision. We also had partially incomplete data on whether a case entered the asylum system as an affirmative case or as a defensive case— missing in the same fashion as the detained/not detained data. We also imputed the median value in the data to these missing data, which amounts to assuming that these cases entered as affirmative cases. Again, we checked to ensure that this assumption did not affect our results, and, as with the assumption about detention status, our results do not change in any substantively meaning way as a result of this assumption. We believe our dataset represents the most comprehensive and innovative on U.S. asylum outcomes to date. As we move through the subsequent chapters we will further introduce the broad range of data we gathered or created to operationalize our theoretically specified models. The data necessary for replication of the models that follow as well as codes for the replication of our results and other papers of interest are available via the authors’ websites.21

CHAPTER 3

A Cognitive Approach to IJ Decision Making

In this chapter we draw upon two disparate disciplines and their literatures. First, law professors whose work has been primarily driven by the expectation that international and U.S. asylum law will be applied consistently and fairly have largely responded normatively to the significant disparity and inconsistency in grant rates within the asylum system. Second, international relations (IR) scholars, whose research is focused on the relationship between states, have primarily focused on the foreign policy dimension of U.S. asylum decisions and the normative and behavioral questions that are raised by connection. These two sets of scholars have largely operated in isolation from each other’s work, while at same time both literatures have failed to draw upon a third body of scholarship that could greatly enhance their understanding of asylum outcomes. We seek to integrate these approaches and to supplement them with a cognitive variation of the attitudinal model from the judicial behavior literature. By incorporating a cognitive model of decision making, we can show how IJs consider some pieces of information objectively while others are treated subjectively. This approach allows us to account for informational cues that influence decisions while assessing the impact of national interests and human rights interests. The influence of these competing factors is conditional on a judge’s policy preferences toward asylum. We find that the policy predispositions of the IJs play a dominant role in explaining the observed discrepancies in asylum grant rates and that liberal IJs respond to certain applicant characteristics differently than their more conservative colleagues—findings that strongly implicate norms of fairness and consistency. At the same time, we also find some limited evidence that the law constrains the decision making of the IJs with respect to applicant characteristics—a finding that speaks to one of the central questions in the study

Cognitive Approach

49

of judicial behavior: to what extent does law constrain the use of preferences? In the following sections, we first frame the issue of asylum outcomes within the broader human rights compliance literature and the dominant debates within the IR literature, and then turn to the judicial behavior literature and place immigration judges (IJs) within that theoretical framework. We then set forth a cognitive model of asylum decision making for IJs and derive a set of empirically testable hypotheses, which we test with two types of statistical models. We find that, true to our attitudinal and cognitive approach, judicial policy preferences strongly predict IJ decision making and that humanitarian factors are treated more objectively than are material interests in the decision-making process.

International Relations Perspective on U.S. Asylum Obligations IR scholars have largely focused on the role of U.S. foreign policy concerns in U.S. refugee policy. As we discussed in the introduction, refugee policy was essentially an indistinguishable component of immigration control under congressional control; however, successive presidential administrations circumvented rigid congressional quotas through their attorney general’s use of parole power. These refugee decisions made by the executive branch were driven primarily by their relevance to foreign policy and national security concerns (Morris 1985, 41–42). Indeed, until the early 1980s, this executive prerogative was rarely challenged as “asylum for refugees was widely viewed as serving foreign policy goals and could be managed so as to assuage the potential for domestic political for domestic political anxiety; and finally, it was assumed that only communist regimes generated refugees” (Gibney 2004b, 150). Not surprisingly, the foreign policy concerns produced obvious double standards that defied international human right standards. In contrast to Cubans, Haitians were treated differently. As Loescher and Scanlan (1986, 79) note, “There was no sense that it was in the U.S. national interest to characterize Haitians as victims of persecution.” The difference was that Cuba’s Castro was a Cold War enemy and Haiti’s successive Duvalier dictators were deemed to be solidly anti-communist. In addition, the United States worried about further destabilizing the country, which would have encouraged even more refugees. Eventually, the disparate treatment of Cuban and Haitian refugees triggered congressional hearings in 1975 and

50

Chapter 3

1976. These hearings identified the role of the Department of State in early asylum claims. As Loescher and Scanlan (1986, 82) point out, what came to light was that “although the INS had always been entrusted with the duty of evaluating individual asylum claims . . . it was required to seek an ‘advisory opinion’ from the State Department on doubtful cases and those which the INS considered to be without substance.” Review was rarely individualized but rather guided by the presumption that Haitian asylum claims were not valid. For Cubans, the presumption was the reverse. Thus the decisions were made “on the basis of considerations having little to do with individual fear of persecution and much to do with relations with countries accused of persecuting” (Loescher and Scanlan 1986, 82). In response, however, there were further reforms in 1990 during the first Bush administration that were “intended to ensure that political-asylum rulings are ‘fair and sensitive’ ” (Koehn 1991, 232). Specifically, in addition to the establishment of the Asylum Officer Corps, a documentation center was established with information on country of origin human rights conditions. Most important, the asylum officers (AOs) were “encouraged to issue rulings that are not dependent upon the State Department’s advisory recommendations” (Koehn 1991, 232). Gibney (2004b, 158) describes the change in even stronger terms: “[T]he process is also required to be ‘independent of State Department prerogatives.’ ” Documentary standards were also altered to allow testimony without corroborating evidence “if the personal account is ‘credible in light of general conditions’ in the country of origin” (Koehn 1991, 232). At least in the case by case determination of eligibility for asylum, the U.S. process had moved much closer to the standards expected in the Refugee Convention. Early systematic empirical evidence generated by IR scholars has demonstrated that throughout the 1980s, U.S. asylum and refugee policy continued to be driven primarily by foreign policy concerns rather than human rights commitment (Gibney 1988; Gibney and Stohl 1988; Gibney, Dalton, and Vockell 1992; Loescher and Scanlan 1986), and subsequent post–Cold War studies have also generally concluded that U.S. economic and material interests influence grants more than human rights conditions (for example, Hassan 2000; Gibney 2004a). Thus, there appears to be a potential gap between U.S. promises under international and domestic law and the law’s actual practice, a gap that is similar to the one found throughout the larger IR compliance literature on human rights treaties (e.g., Keith 1999; Hathaway 2002; Hafner-Burton and Tsutsui 2005).

Cognitive Approach

51

The empirical work that has sought to understand the (dis)connection between commitment and compliance in regard to international human rights treaties has largely turned to two competing sets of theoretical explanations within the field of IR—realism and constructivism. Realists perceive states as rational actors whose behavior is based primarily upon narrow selfinterest (Waltz 1979; Keohane 1984). State commitments to international humanitarian or rights norms are perceived as “cheap talk” (Mearsheimer 1994) that gives way to more substantive state interests. Conversely, constructivists emphasize the emergence and diff usion of international human rights norms through networks of domestic and transnational actors, who not only shape the discourse of international human rights but also rally publics to convince states to formally accept and adhere to these norms (e.g., Keck and Sikkink 1998; Risse, Ropp, and Sikkink 1999). In addition, states are expected to comply with these norms because states have a propensity to comply with their legal obligations (Henkin 1977) or because they generally aspire to comply with norm of pacta sunt servanda (agreements must be kept) (Chayes and Chayes 1993). However, the world society approach suggests the possibility of a decoupling effect between treaty commitment and practice (see Hafner-Burton and Tsutsui 2005). This approach (e.g., Meyer et al. 1997) perceives states to be embedded in an integrated cultural system that “promulgates cognitive frames and normative prescriptions that constitute the legitimate identities, structures, and purposes of modern nationstates” (Cole 2005, 477). Thus, with the proliferation of human rights treaties codifying human rights norms, states’ legitimacy or “good nation” identity is increasingly linked to the formal acknowledgment of these norms (Cole 2005; Wotipka and Ramirez 2007); however, as Cole (2005, 477) notes, many states join the traditionally weak human rights regime, “not out of deep commitment, but because it signals their probity to the international community” and thus “a decoupling is endemic to the human rights regime.” This norms versus interest debate is reflected in IR-based asylum literature, although the link to treaty compliance and that empirical literature is typically overlooked (Rosenblum and Salehyan 2004; Salehyan and Rosenblum 2008; Rottman, Fariss, and Poe 2009; Keith and Holmes 2009; Holmes and Keith 2010; Keith, Holmes, and Miller 2013). More recently the compliance literature has broadened beyond IR theory and has begun to consider the role of domestic institutions. The domestic institutions approach dismisses the assumption of a unitary state actor and instead recognizes the role of numerous domestic actors and institutions

52

Chapter 3

within the state (particularly in democratic regimes), which may affect the regime’s calculation of costs related to commitment and (non)compliance. Democratic electoral processes and legal institutions are seen as providing the public and other political actors with the tools and venues through which they can hold the regime accountable should it fail to keep its international commitments (Keith 2002, 2012; Neumayer 2005; Hathaway 2005; Powell and Staton 2009). From a broader perspective, democratic affinity for the rule of law and respect for constitutional constraints and judicial processes in the domestic context arguably carry over to the international context and increase the likelihood that democratic regimes will honor their international legal commitments (Simmons 2000). Assumptions of the domestic institutions approach have received much more specific empirical attention than those of realist and rationalist theories, presumably because scholars can more readily observe and measure domestic institutional contexts than costbenefit calculations or the diffusion of norms. However, as far as we know, no study has examined the actual behavior of actors on the ground implementing the treaty commitments. Our contribution to this literature is to pull the focus to a court-level bureaucrat—the IJ. The IJ primarily enforces one of the core components of U.S. commitment under the Refugee Convention— the granting of asylum and the protection of the norm of non-refoulement. To provide a foundation for our contribution we begin with the work of Rosenblum and Salehyan (2004). They shift the focus away from the either/ or assumption in the norms versus interest debate and instead more accurately examine the relative importance of these two sets of values (presumably to U.S. foreign policy makers), recognizing the potential for them to complement as well as contradict each other. These authors broadly define U.S. interests to include diplomatic, security, economic, and domestic demands (such as migration control) and define normative goals to be those that “reflect a truly humanitarian, even altruistic concern with the needs of asylum applicants, regardless of U.S. relations with their country of origin” (681). They then model the values in two-dimensional space, with humanitarian norms representing one axis and interests representing the other axis, creating four quadrants representing the value(s) potentially triggered by the asylum seeker’s country of origin.1 The most theoretically informative categories for the norms versus interest debate are those in which humanitarian concerns and interests would predict contradictory outcomes: normative interests predict grant but U.S. interests predict denial or U.S. interests support a grant but normative interests are lacking.

Cognitive Approach

53

Rosenblum and Salehyan argue that the outcome depends upon the relative weight of the humanitarian and interest-based values, which we take as our point of departure, and thus we make two significant shifts away from this literature. First, we argue that to understand the role of these values in U.S. asylum outcomes, it is necessary to shift to the micro level of analysis and to examine the individual decision makers—IJs—and their decisionmaking processes. Second, we argue that the nature and context of asylum decision making shape the processes IJs rely upon to reach decisions, and ultimately we argue that the approach they employ for objective legal considerations is different from the one through which they process the more abstract extralegal factors related to national interests. Thus “norms” and “interests” are processed differently by these decision makers. We turn next to this discussion of IJs and the institutional context in which they make asylum decisions.

The Institutional Context of Immigration Judge Decision Making IJs arguably have less formal independence than federal judges and potentially less independence than administrative law judges. Nonetheless, they maintain a high degree of independence. The Immigration and Nationality Act (Section 240) states that “in deciding the individual cases before them . . . IJs shall exercise their independent judgment and discretion.” Despite their lack of life tenure, IJs experience a good deal of autonomy in their decision making for several reasons. First, they decide a large volume of cases, approximately three times the number of cases decided by a typical federal district court judge. For example, in 2009, 250 IJs were tasked with deciding approximately 350,000 cases, or 1,400 cases per judge per year, compared to 400 cases per year per district court judges in the same year. There is also a growing case backlog.2 Such a tremendous caseload makes monitoring IJs’ behavior difficult.3 Second, IJs face a low probability of being overturned by the Board of Immigration Appeals (BIA). Between 1990 and 2010 about 12 percent of all IJ asylum decisions were reversed. Third, the standard of appellate review on facts or for matters of discretion to which the IJs are subjected is exceedingly low—clearly erroneous. Thus, the BIA would have to find that the IJ reached unreasonable conclusions in order to reverse a decision, whereas application of the law is reviewed de novo. In essence, we conclude

54

Chapter 3

that IJs are judges-as-bureaucrats, with ample discretion and broad civil ser vice protections (see, e.g., Brehm and Gates 1997). The legal strictures in asylum cases are loose because both the facts and the law are vague (Baum 2010; Legomsky 2010). Law (2005, 830) notes “the indeterminacy of the governing legal standards” in asylum cases, which leaves judges “to define vague yet crucial terms—‘political,’ ‘persecution,’ ‘well-founded fear,’ ‘more likely than not’—on a case-by-case basis” with “precedent provid[ing] only limited guidance, given the dependence of asylum claims on case-specific facts.” As Martin (2000, 3) observes, the “basic facts in any particular [asylum] case are highly elusive” and “the adjudicator has to decide what happened in a distant country” with only two imperfect sources: general human rights country reports and the personal testimony of the asylum seeker. Alexander (2006, 19) notes that immigration courts routinely lack evidence: “[T]he witnesses, objects, and documents that could prove or disprove a fear of persecution, for example, are likely beyond reach oversees” and “indeed, the ability to gather evidence may be blocked by the very government alleged to be the persecutor.” In fact, IJs have been criticized by Federal Court of Appeals judges for an exaggeration of the availability of evidence in many countries of origin (for example, see Judge Posner in the Seventh Circuit opinion in Iao v. Gonzales, 400 F.3d 530, 533–35). Further compounding factors include the reality that over three-fourths of the applicants do not speak English (any one of 389 languages may be spoken at the hearings) and that most applicants in typical immigration proceedings are not represented by counsel, requiring judges to guide unrepresented individuals through the proceedings, advise them of their rights and options for relief of removal, and answer questions (Alexander 2006; TRAC 2009). Finally, the IJs have to deal with their heavy caseload with little staff assistance.4 IJs complain about the “the constant drumbeat of case completion goals,” that making credibility determinations “is extremely, extremely difficult,” and most important that “there is not enough time to do research and adequately read about country conditions” (Lustig et al. 2008, 65– 66). Clearly, these are overburdened and under-resourced courts with high stakes for applicants. As we discussed in the introduction, the most frequent criticism of the IJs has been the disparities in grants and denials among them (e.g., TRAC 2006; Ramji-Nogales, Schoenholtz, and Schrag 2009), which suggest the influence of the judges’ policy preferences or personal biases. Law professors

Cognitive Approach

55

argue that subjects such as asylum, which “inspire ideological or emotional fervor would seem to have the greatest disparate outcomes, since the fleshand-blood adjudicators who decide the cases will have extra reason to resolve the more indeterminate questions by resort to visceral beliefs and emotional impulses” (Legomsky 2007, 442). Law (2005, 830) specifically notes that asylum cases invoke highly charged political issues and that asylum is “an area of law that divides the Ninth Circuit along ideological lines, a fact acknowledged by the judges themselves and borne out by the data.” Furthermore, Law (2010) and Law and Williams (2012) have found that judges on the Third, Fift h, and Ninth Circuit Courts tend to decide cases in a manner that is congruent with their ideological preferences. The nature and the context of these decisions lead to a reliance on policy preferences by IJs in decision making. The lack of certainty in legal standards, the frequent lack of corroborating evidence, and the complexity of credibility determinations invite the use of motivated reasoning in which directional (or policy-motivated) goals overwhelm accuracy goals (Braman and Nelson 2007). The prominence of these directional motivations is important to our cognitive approach to judging because these directional or policy goals will tend to induce what Bartels (2010) has termed top-down decision making. In a top-down approach, the judge brings a theory or predisposition to the case—this is analogous to the ideological proclivities that U.S. Supreme Court Justices bring to cases. This top-down approach is contrasted with a bottom-up approach in which facts are evaluated deliberately and decision making is constrained by the manner in which the law dictates the treatment of those facts. We develop these concepts further in the following section. An additional contextual factor—the overwhelming workload of IJs— further induces reliance on policy predispositions to decide cases and likely increases IJs’ reliance on informational cues to decide cases (Baum 2010). Thus, rational litigant behavior is difficult because of the unpredictability of the BIA on appeal, since it will be difficult to predict in any given case how the reviewing authority would treat cases with highly ambiguous facts and few legal constraints. Furthermore, as we discuss at length in Chapter 5, though applicants—who overwhelmingly drive the BIA’s agenda—are rational, they are not rational in a manner that engenders hierarchical control because they do not appeal those cases that the BIA is, necessarily, most likely to wish to reverse. Th is means that institutional control of these adjudicators is diminished, since rational litigant behavior is thought to be a necessary

56

Chapter 3

precondition for bringing cases deviating from the principal’s preferences that will allow it to control agents lower in the hierarchy (Songer, Cameron, and Segal 1995). We turn to the judicial behavior literature to further explore these issues.

The Judicial Behavior Literature and Judging in Asylum Cases Scholars of judicial behavior have tended to focus their attention on three interrelated sets of explanations for judicial choice: attitudinal, strategic, and legal models. In brief, the attitudinal model holds that judges primarily pursue their policy preferences (e.g., Segal and Spaeth 2002), the strategic model holds that judges will pursue their policy preferences in light of the preferences of other important actors, such as other judges on the same court, the court above them in the judicial hierarchy, or other relevant institutions (e.g., Epstein and Knight 1998), and the legal model assumes that judges’ choices are constrained by the dictates of precedent, the plain meaning of statutes, and the intent of the framers of statutes (e.g., Gillman 2001). Of these models, we focus on the attitudinal model for reasons explained above and explored further below. To this explanation, however, we add an explicitly cognitive approach. In essence, we posit that the policy preferences of IJs influence their decisions in asylum cases, but that U.S. asylum law also imposes some constraints on the use of the policy proclivities of an IJ. We believe that a comprehensive model of IJ decision making approximates a situation in which an IJ, under tremendous time pressure and unsure of the credibility of an asylum seeker, will use policy predispositions to help process both legally relevant and legally irrelevant facts. The cognitive approach is operationalized in our model as the differential weights placed by an IJ on particular facts, which we believe will vary with the policy predispositions of the judge. A similar approach has been usefully summarized by Bartels as a mix of top-down and bottom-up approaches.5 After summarizing the attitudinal aspects of asylum decision making, we integrate an attitudinal approach with this cognitive process. As noted above, asylum decision making is likely to activate an ideological schema. The judicial behavior literature portrays judicial votes on rights claims as a liberal-conservative dichotomy, with support of rights claims considered to be the liberal position and support of the government interests

Cognitive Approach

57

over the rights claims to be conservative (e.g., Segal and Spaeth 2002). This dichotomy has been applied to anti-alien/pro-alien behavior in federal immigration cases (e.g., Law 2005, 2010; Westerland 2009; Ocepek and Fetzer 2010; Williams and Law 2012). Following the asylum and judicial behavior literatures, we believe that the ideological divide on asylum issues parallels that of other immigration issues, with the anti-asylum viewpoint reflecting more closely conservative or right-leaning perspective and the pro-asylum viewpoint reflecting more liberal or left-leaning perspective of both U.S. and European politics. Therefore, judging these cases with policy proclivities in mind is relatively easy, since asylum cases are likely to activate predetermined schema. Moreover, as discussed earlier, IJs have much discretion in their decision making, given the voluminous caseload, low likelihood of appellate reversal, and high threshold of the reasonableness standard. We believe, given these facets of decision making in the U.S. asylum process, that IJs’ policy predispositions will be highly determinative of their likelihood of granting asylum. More specifically, we derive the following primary hypothesis: Hypothesis 1: IJs with more liberal policy preferences will be more likely to grant asylum than will IJs with conservative policy preferences. The applicability of strategic and legal approaches must be assessed for asylum decision making. Strategic decision making is less likely in asylum cases than it is in other kinds of decision contexts. Though a relatively large number of asylum cases are appealed, this creates a very high workload for reviewing members of the BIA, meaning that any given case is likely to receive only cursory review. In addition, the difficulty of predicting what the reviewing authority will do with ambiguous facts in any specific case inhibits hierarchical control of the IJs (we explore the rationality of litigants in asylum cases further in Chapter 5). Moreover, IJs do not decide cases in panels and therefore have little to fear from other IJs in terms of review. The vagueness and ambiguity of asylum cases also lesson applicability of the legal model to the decision making of IJs (Baum 2010; Legomsky 2010), although we do find some evidence that IJs make decisions that reflect the assessment of legally relevant factors. We know from the literature that asylum grant rates are influenced by extralegal factors, such as U.S. military aid or trade relations (Rosenblum

58

Chapter 3

and Salehyan 2004; Rottman, Fariss, and Poe 2009), or potential indicators of a bogus asylum seeker (Keith and Holmes 2009). Neumayer (2005) finds divergent recognition rates for applicants of the same nationality in the EU and doubts that the differences are due to individual case factors. The work of Bohmer and Schuman (2008) suggests that asylum adjudicators in both the United States and the United Kingdom perceive evidence and facts through limited cultural filters. In addition, the disparities in grant and denial rates among judges, even within the same court (e.g., TRAC 2006; Ramji-Nogales, Schoenholtz, and Schrag 2007, 2009), suggest the influence of the individual judge’s policy preferences or personal biases. Our expectation is that these extralegal factors will tend to be emphasized differently depending on an IJ’s policy predisposition. As noted above, we adopt a cognitive approach to decision making in asylum cases to explain this variation. Essentially our approach is that IJs use a mixed process—in which their policy predispositions are moderated in certain circumstances but not in others. Bartels (2010) describes this type of process as a continuum ranging from totally top-down decision making in which the predispositions that a judge brings to a case predominate any new information they encounter to totally bottom-up decision making in which a decision maker objectively scrutinizes the facts before him or her. We believe that the “norms” factors, such as the level of human rights abuse reported in a country, are likely to be evaluated in a more bottom-up fashion (i.e., somewhat objectively),6 whereas material and security concerns, such as whether the applicant is from a country producing high numbers of illegal immigrants, are likely to be evaluated in a more top-down fashion (i.e., somewhat subjectively). Put differently, IJs will assess human rights conditions in a manner that is less contingent on their own policy preferences than on their assessment of material and security concerns, which will be more contingent on their policy predispositions. Specifically, we derive the following hypothesis in regard to the cognitive model. Hypothesis 2: The effect of extralegal factors, such as material and security interests, will be more influenced by the policy preferences of the IJs than will be more objective legal considerations, such as human rights conditions. This hypothesis is premised on three important facts about IJ decision making. First, asylum law dictates that IJs should consider the likelihood of

Cognitive Approach

59

persecution should an applicant be returned to a country (Legomsky 2010). Second, given the severe time constraints under which IJs make decisions, their search for information about a par ticu lar applicant is limited (Lustig et al. 2008). Therefore, the IJ in a case is heavily reliant on U.S. Department of State reports about an applicant’s country, and these reports are almost solely concerned with the human rights situation in a country (Bohmer and Shuman 2008). Thus, three factors lead us to conclude that human rights conditions should be evaluated in a more bottom-up fashion than material or security concerns: (1) the law dictates consideration of human rights conditions, (2) evidence of these conditions is also some of the most readily available evidence in a case, and (3) evidence of these conditions is in fact often the only available evidence, other than the applicant’s testimony. In addition, extralegal factors such as the level of trade between the applicant’s country and the United States—the type of thing that the literature refers to as a material interest—are not presented to IJs and are not supposed to be considered in determining an asylum application. Therefore, these types of factors are likely to be evaluated in a manner that is closer to the top-down end of the decision-making continuum—that is, treatment of this type of variable should be dependent on a judge’s policy proclivities. Third, IJs are trained by the Executive Office for Immigration Review (EOIR) to evaluate and make credibility determinations with respect to the likelihood of persecution should an applicant be returned to his or her country of origin, but they are not trained to make their determinations on the basis of any other factors (see U.S. Department of Justice, Executive Office of Immigration Review 2009). Psychological approaches to decision making indicate that two interrelated factors will tend to make decision makers approach par ticu lar facts from the bottom up: accountability and fear of invalidity. When decision makers must justify their decisions to someone else (when they are accountable), they tend to be more objective in their decision making (Lerner and Tetlock 1999). In addition, when a decision maker must fear invalidity (in this case being overturned), they tend to take a more bottom-up approach to their decision making (Fazio and Towles-Schwen 1999). Importantly, because review of an IJ’s decision is to be premised on asylum law, and because asylum law dictates consideration of certain facts but not others, these two factors, accountability and fear of invalidity, should work to make evaluation of human rights concerns more objective than other factors. To summarize, we have two major hypotheses about the decision making of IJs in asylum cases. First, asylum implicates clearly identifiable and

60

Chapter 3

emotionally compelling policy choices. Given the fact that reversal of IJ decisions is unlikely, and that when it does occur the outcome of review will not be predictable, we hypothesize that IJs will use their policy predispositions to guide their decision making. Second, we hypothesize that the assessment of human rights concerns will be relatively objective, while the assessment of material and security interests will be more subjective than the treatment of humanitarian factors. Together, we show that these two hypotheses explain a good deal of the observed variation in IJ decision making.

Analysis Our focus throughout this chapter is on two complementary questions that are central to the debate on judicial decision making in the social science (primarily political science) and legal academic communities. First, what role do judicial policy preferences play in determining judicial choices? Second, does the law constrain these choices in any meaningful manner? To get at the first question, we constructed a measure of the policy preferences of IJs in asylum cases, which we termed asylum liberalism, and we presented the results of this process in Chapter 2. As a reminder, asylum liberalism is a proxy variable for IJ ideology that is premised on certain pertinent background characteristics of the IJs. Such a measure is congruent with both attitudinal and cognitive approaches to understanding judicial decision making. According to one leading proponent of the cognitive approach, “All mental processing draws closely from one’s background knowledge. . . . There is a natural coherence between specific inferences and the decisionmaker’s background belief system” (Simon 2004, 536). Asylum liberalism is now the main variable of interest in the models that follow. Our first hypothesis posits that liberalism will be statistically and substantively significant positive predictor of the likelihood that an IJ will grant asylum. To answer the second question, whether the law constrains the use of these policy preferences, we enter a series of interaction terms into the model in an effort to determine whether legally relevant facts (such as the human rights conditions in a country) limit the extent to which asylum liberalism affects the decision making in asylum cases. We create interactions not only for the human rights conditions within a country but also for the material and security interests that IJs and the U.S. government may

Cognitive Approach

61

perceive as being relevant to the relationship between the United States and the home country of the asylee. If the cognitive portion of our story is true, then we would expect that the human rights facts constrain the use of policy preferences more than do material or security concerns. Below we describe in more detail the variables we enter into the analyses—basic descriptive statistics on these variables are available in the online appendix to this chapter, which can be found on any of the authors’ websites.7 Our approach to modeling the data is premised on our theory that judging in asylum cases can best be understood as a combination of attitudinal and cognitive approaches to judicial decision making. We use two types of statistical models to understand the data. Readers who wish to know more about those models than is presented in the text may find details in the online appendix to this chapter;8 those who are not interested in such details may skip the appendix without any loss of understanding of the substantive issues raised by our results. One approach involves directly modeling the dependency in our data that arises because IJs decide multiple decisions within our dataset. By dependency we are referring to the likelihood that each IJ makes decisions that are similar to one another, and that are more similar to one another than they are similar to the decisions of a different IJ. Thus we need a method that allows us to model this dependency. One possible approach is to estimate a random intercept for the IJs in our data. This approach allows the overall propensity to grant relief to vary randomly among the judges, accounting for any underlying part of their likelihood to grant relief that we have not considered in the model (Rabe-Hesketh and Skrondal 2012; Snijders and Bosker 1999). For these random intercept models we use the traditional outcome coding in the asylum literature of relief versus no relief.9 We estimate and present two versions of this model below: one on all of the available data between 1990 and 2010 and one on a restricted subset of the data between 1999 and 2010. The second, restricted, version of the model is estimated for comparison with our second modeling strategy. As we emphasized in Chapter 2, the traditional approach to coding outcomes in asylum cases is to code whether or not an applicant received any form of relief (either type of withholding or asylum). We believe, however, that there is room for a more nuanced approach: treating the four potential outcomes as ordered, based on the substantive benefits conferred under each type of relief. We presented preliminary evidence in Chapter 2 that this substantive ordering is as follows: no relief, withholding under the CAT,

62

Chapter 3

withholding, and asylum. Here we present the full results of that modeling process, known as stereotype logit. It is necessary, however, in using this approach to restrict our sample to the range 1999 to 2010 for two reasons. First, withholding of removal is not used much until the implementation of IIRIRA in the late 1990s. Second, withholding of removal under the CAT is not an option under the law until 1999. For these reasons we estimate the stereotype logit model on data from 1999 to 2010 and present a comparison version of our random intercept model from the same period so that readers may see how our understanding of asylum changes when the dependent variable is reconceptualized. To summarize, we estimate three separate models on two different sets of data using two different modeling approaches. It is notable that, across these modeling choices and periods, our results are stable.

Variable Description Because we have addressed it in Chapter 2, we do not repeat our discussion of asylum liberalism here, except to note that it ranges from 0 to 4, with higher values indicating a background that suggests the IJ will be more permissive in the granting of relief in asylum hearings. We include two measures of the human rights conditions within a country. To assess the level of human rights abuse in the sending country, we employ Gibney’s five-point measure of state-level personal integrity rights abuse, based specifically on the Department of State human rights country reports. Gibney’s is one of the standard measures of political repression and has been utilized in the empirical human rights literature for over three decades (for an exhaustive list, see http://www.politicalterrorscale.org /bibliography.php and Wood and Gibney 2010).10 The abuses captured in this scale are arguably the most egregious forms of human rights abuse (Poe and Tate 1994; Keith 2012). The categories within the scale are as follows: 1. Countries under a secure rule of law, people not imprisoned for their views, torture is rare or exceptional. . . . Political murders extremely rare. 2. A limited amount of imprisonment for nonviolent political activity. Few persons affected, torture and beating exceptional. . . . Political murders rare.

Cognitive Approach

63

3. Extensive or recent history of extensive political imprisonment. Execution or other political murders and brutality common. Unlimited detention for political views accepted. 4. Practices of level 2 expanded to larger numbers. Murders, disappearances are common, but terror affects primarily those who interest themselves in politics or ideas. 5. Terrors of level 3 expanded to the whole population. No limits on means or thoroughness with which leaders pursue personal or ideological goals. (Gastil 1980) In addition, we follow the asylum literature and include a measure of institutionalized democracy (Polity IV’s eleven-point additive scale; Marshall and Jaggers 2009)—which we refer to as democracy—on the assumption IJs will tend to associate the threat of persecution with authoritarian regimes and may be more skeptical of asylum claims from highly democratic countries. Per the recommendation of the scale’s creators, we recode the Polity eleven-point measure into a trichotomous regime measure, where 0 represents autocracy, 1 represents anocracy, and 2 represents democracy (see http://www.systemicpeace.org/polity/polity4.htm). As noted, we interact each of these variables with asylum liberalism to capture the cognitive aspects of our theory of IJ decision making. Rosenblum and Salehyan (2004, 681) delineate U.S. instrumental interests as those that seek “to preserve good relations with allies (i.e., by rejecting their refugees), to weaken opponents (i.e., by accepting their refugees), and to limit ‘back-door’ access to the United States through illegitimate asylum claims.” Following Rosenblum and Salehyan, we include four indicators to capture the material and security interests of the United States vis-à-vis a country of origin. To capture the security or strategic relationship between the United States and the country of origin, we utilize two measures. We code the bilateral trade between the United States and a sending nation in a given year, which we then log to account for skew. Our expectation is that high levels of trade between a sending nation and the United States will decrease the likelihood of relief being granted. We include a dummy variable equal to 1 when the country of origin is a recipient of military aid from the United States. Again, we expect that when military assistance is given to a country the United States will be less likely to grant those fleeing that country asylum.

64

Chapter 3

To account for the possibility that the asylee might be perceived as a so-called “bogus asylum seeker,” we include two variables. One variable is whether the applicant comes from one of the ten top source countries of illegal immigrants. Top ten illegal immigration is a dummy variable equal to 1 when the country of origin is one of the top ten producers of illegal immigrants and 0 otherwise. Our expectation is that hailing from one of these countries will decrease the likelihood of receiving relief from an IJ. While Rosenblum and Salehyan used only the undocumented immigrant sending country as an indicator of a potential bogus asylum seeker, our previous work has demonstrated the level of economic development of the sending country may also serve as such a cue (Keith and Holmes 2009; Holmes and Keith 2010; Keith, Holmes and Miller 2013). Thus we include a second variable, World Bank development, which is ordinal and ranges from 0 to 3, with higher scores attached to applicants coming from countries that are more developed economically. We expect that those coming from more developed countries will have an easier time gaining relief, given that they are potentially less likely to be economic migrants. We interact asylum liberalism with each of the material and security interest variables, and our expectation is that these factors will not constrain attitudinal decision making to the extent that the human rights factors will. In addition to these main variables of interest we include a host of control variables that are motivated by the empirical literature on asylum in the United States. With respect to the IJ, we include the dummy variable woman, which is equal to 1 if the judge is a woman and 0 otherwise. It is important to control for gender as others have found that gender is an important predictor of the granting of relief (e.g., Ramji-Nogales, Schoenholtz, and Schrag 2007, 2009). In previous empirical investigations, women have been significantly more likely to grant relief, and so we expect this variable to be positive. Songer and Crews-Meyer (2000, 759) suggest that “female judges may be quicker to empathize with underdogs in a variety of civil liberties issues since these judges have either experienced or witnessed the problems involved in being a political minority.” Ramji-Nogales, Schoenholtz, and Schrag (2009, 47–48) note that such experiences “might make female judges more sympathetic to stories of persecution, as well as more conscious in eliminating their own biases from the decision making process.” They also suggest that the gender effect may reflect greater empathy toward the applicant and a less combative setting that may result in more coherent testimony.

Cognitive Approach

65

In addition, we include two variables that capture the language spoken by the applicant. First, English is equal to 1 if the applicant speaks English and 0 otherwise. We expect that English speakers may be more likely to get relief because they can better understand the process and advocate for themselves and because it may be that IJs see them as less likely to be a potential burden on the state (Rottman, Fariss, and Poe 2009; Keith and Holmes 2009; Holmes and Keith 2010). Second, Arabic is equal to 1 if the applicant speaks Arabic and 0 otherwise. Our expectations with respect to Arabic speakers are ambiguous because, on the one hand, it may be that Arabic speakers are more likely to be suspected of potential terrorist connections (making them less likely to receive relief), but on the other hand IJs may see some strategic benefit accruing to the United States by granting relief to those fleeing regimes in countries in which Arabic is spoken predominantly (see Holmes and Keith 2010). In addition to the language spoken by the applicant, we include three additional variables that are attached specifically to the applicant. Legal representation is equal to 1 if the applicant is represented by an attorney or other advocate and is coded 0 otherwise. We expect that those who are represented will fare considerably better in their asylum proceedings than those who are not represented, consistent with previous work in this area (RamjiNogales, Schoenholtz, and Schrag 2007). Affirmative application is equal to 1 if the applicant is proceeding affi rmatively in the process (and 0 otherwise)—that is, this variable is equal to 1 if the alien initiated the asylum proceedings as opposed to filing for asylum defensively to protect himself or herself against deportation or removal. This variable should be strongly associated with an increased likelihood of receiving relief. We also control for the detention status of the asylee. We expect that this variable will have a strong negative impact on the likelihood of receiving relief, as those in detention or who have been previously detained will have a more difficult time advocating on their own behalf (Ramji-Nogales, Schoenholtz, and Schrag 2007) and may be viewed with more suspicion by IJs. We code detention status as a trichotomous variable equal to 0 if the applicant has never been detained, 1 if the applicant was previously detained but has been released at the time of the hearing, and 2 if the applicant is detained at the time of the asylum hearing. The models include five variables meant to capture potential change in the policy context within which IJs decide asylum claims. First, we expect

66

Chapter 3

that the partisanship of the attorney general might affect IJs’ decision making if they behave strategically. That is, when the attorney general is a Democrat, as measured by the Democratic administration variable, IJs might be more likely to grant relief to applicants. Second, we include the variable Nine Eleven, equal to 1 in September 2001 and 0 otherwise, to control for any temporary effect the attacks of that day may have had on the likelihood of being granted relief. In addition, we include three variables that measure major policy changes in U.S. asylum law. Two of these variables capture changes introduced by IIRIRA in the mid-1990s. First, IIRIRA expedited removal is equal to 1 after April 1997 and 0 before. This variable captures the implementation of expedited removal procedures for aliens arriving at U.S. ports of entry. Second, IIRIRA one-year ban is equal to 1 after April 1998 and 0 before. This variable captures the effect of the implementation of the one-year deadline for the filing of an asylum application. In essence, under this provision of IIRIRA, applicants are required to fi le their application for asylum within one year of their arrival or face removal (unless exceptional circumstances are found to explain the delay in filing). We expect both of these variables to increase the overall grant rate for IJs. Although we explore why this is so in great detail in subsequent chapters, a bit of explanation is warranted. In essence, the effect of both of these provisions, conditional on an applicant reaching the IJ, is to increase the overall strength of the applicant pool, as many applicants have already been removed before this stage. Finally, we include a variable, Real ID, that is equal to 1 after May of 2005 and 0 before. The scholarly literature on asylum is split on the likely impact of the Real ID Act, which potentially made credibility determinations by IJs more subjective (Cianciarulo 2006) or simply codified existing case law (Anker 2011). To the extent that Real ID made reaching IJs more difficult, our expectation for Real ID is the same as it is for the IIRIRA interventions: we expect that the implementation of Real ID will increase the grant rate of IJs. We treat these statutes as control variables in the analyses of this chapter; in Chapter 6 we explore them further in the broader policy context. We include a variable that captures the economic situation in the United States when an asylum hearing occurs. Although we will explore how local economic and political conditions affect IJ decision making in Chapter 4, here we simply wish to control for how national economic trends might alter IJ decision making without overcomplicating an already complex model. To this end, we include national unemployment (lagged one month). Some

Cognitive Approach

67

evidence suggests that concern over immigration is primarily premised on concern over what immigrants will do to the labor market (i.e., taking jobs from citizens) (Citrin et al. 1997; Cornelius and Rosenblum 2005), and we expect that as unemployment rises the likelihood of getting relief will decrease. Finally, we include a count variable of the number of months elapsed from the beginning of our dataset, which we call elapsed time. We include this time trend variable because there is a strong upward trend in grant rates from 1990 to 2010. Th is trend is worth exploring in its own right, and we undertake a time series analysis in Chapter 6 that explores fluctuations in the number of applicants granted relief over time that are otherwise not captured in the model.11 Here we simply wish to control for any potential confounding effects that accrue over time. Descriptive statistics, as well as information on the collection of each of the variables, is available in the accompanying online appendix to this chapter.

Results Table 3.1 presents the effects of each of the variables derived from the models that are presented in full in this chapter’s online appendix. We also present a simple model without interactions in the online appendix to allow for more direct interpretation of the interacted coefficients. The first column in the table presents predicted probabilities from the random intercept model using all of the available data (1990–2010), the second column presents predicted probabilities from the random intercept using just the data from 1999 to 2010, and the third, fourth, and fift h columns present the results from the stereotype logit model for the various potential outcomes (we do not present results for the possibility of withholding under the CAT to save space and because this possibility remains relatively unlikely). First, we note the consistent effect of asylum liberalism across the models. Increasing the liberalism of the IJ from the 5th to the 95th percentile increases the likelihood of a grant of relief (or asylum) by between 13 and 27 percentage points. Because of the interactions with asylum liberalism, some clarification is warranted in interpreting the asylum liberalism term. The changes in predicted probability presented for this variable represent the effect of asylum liberalism when all of the material and security interests and the human rights conditions are set at their means or medians. The effect

Table 3.1. Predicted Probabilities of a Grant of Relief Variable Judicial policy preference Asylum liberalism (+)

Random Intercept; 1990−2010

Random Intercept; 1999−2010

0.13 [0.08, 0.17]

0.27 [0.22, 0.31]

U.S. material & security interests Log of trade with −0.22 [−0.23, −0.22] U.S. (−) U.S. military aid (−) −0.02 [−0.03, −0.02] Top ten illegal −0.19 [−0.20, −0.18] immigration (−) World Bank develop0.06 [0.06, 0.07] ment class (+) Human rights conditions Democracy (Polity) (−) −0.12 [−0.12, −0.12] Human rights abuse 0.20 [0.19, 0.20] (PTS-State Dept.) (+) Interaction terms Log of trade × asylum liberalism Military aid × asylum liberalism Top ten illegal × asylum liberalism

Stereotype Logit; No Relief −0.18 [−0.29, −0.07]

Stereotype Logit; Withholding

Stereotype Logit; Asylum

0.02 [0.01, 0.02]

0.17 [0.07, 0.27]

−0.22 [0.21, 0.22]

0.16 [0.15, 0.16]

−0.01 [−0.01, −0.02]

−0.14 [−0.14, −0.14]

— −0.21 [−0.22, −0.20]

— 0.26 [0.21, 0.31]

— −0.02 [−0.03, −0.02]

— −0.24 [−0.28, −0.19]

—

—

—

0.18 [0.18, 0.19] −0.17 [−0.18, −0.16]

−0.02 [−0.01, −0.02] 0.02 [0.02, 0.02]

−0.18 [−0.18, −0.17] 0.16 [0.15, 0.16]

0.07 [0.07, 0.08]

−0.14 [−0.14, −0.15] 0.25 [0.24, 0.25]

see Figure 3.1

see Figure 3.1

—

—

see Figure 3.1

see Table 3.2

see Table 3.2

—

—

see Table 3.2

see Table 3.2

see Table 3.2

—

—

see Table 3.2

World Bank development × asylum liberalism Democracy × immigration liberalism Human rights abuse × immigration liberalism Controls Judge woman (+) English speaker (+) Arabic speaker (+/−) Legal representation (+) National unemployment (one-month lag) (−) Democratic administration (+) IIRIRA expedited removal IIRIRA one-year bar Real ID Nine Eleven (−) Affirmative application (+) Detention status (−) Elapsed time (+)

see Figure 3.2

see Figure 3.2

—

—

see Figure 3.2

see Figure 3.3

see Figure 3.3

—

—

see Figure 3.3

see Figure 3.4

see Figure 3.4

—

—

see Figure 3.4

−0.09 [−0.14, −0.03] −0.02 [−0.04, −0.00] — −0.06 [−0.07, −0.04] 0.07 [0.06, 0.08]

0.01 [0.00, 0.01] 0.00 [0.00, 0.00] — 0.01 [0.00, 0.01] −0.01 [−0.01, −0.01]

0.08 [0.03, 0.13] 0.02 [0.00, 0.04] — 0.05 [0.04, 0.07] −0.06 [−0.07, −0.06]

−0.03 [−0.04, −0.03]

0.00 [0.00, 0.00]

0.04 [0.03, 0.04]

0.05 [0.02, 0.09] 0.02 [0.02, 0.03] 0.02 [0.01, 0.02] 0.05 [0.05, 0.06] — −0.01 [−0.00, −0.01] 0.04 [0.03, 0.05]

0.07 [0.03, 0.12] 0.04 [0.03, 0.04] 0.03 [0.02, 0.04] 0.06 [0.06, 0.07] −0.08 [−0.09, −0.08]

0.04 [0.04, 0.04] —

—

—

—

0.02 [0.01, 0.03] 0.09 [0.08, 0.09] — 0.03 [0.02, 0.03]

— 0.04 [0.03, 0.04] 0.04 [0.02, 0.06] 0.05 [0.04, 0.05]

— −0.04 [−0.05, −0.02] — −0.05 [−0.06, −0.04]

— 0.00 [0.00, 0.01] — 0.00 [0.00, 0.00]

— 0.03 [0.02, 0.05] — 0.05 [0.03, 0.06]

−0.12 [−0.13, −0.12] 0.39 [0.39, 0.40]

−0.15 [−15, −0.14] 0.27 [0.26, 0.28]

0.15 [0.15, 0.16] −0.25 [−0.26, −0.24]

−0.01 [−0.01, −0.01] 0.02 [0.02, 0.02]

−0.14 [−0.14, −0.13] 0.23 [0.23, 0.24]

70

Chapter 3

of IJ ideology appears to be more pronounced later in the period: in the full model the “average” effect for IJ ideology is 13 percentage points, but in the model with data from 1999 forward the effect is 27 percentage points. It is worth noting that, in this later time frame, the effect of IJ ideology is (with a counter for elapsed time) the most important single predictor of outcomes for applicants, mattering more than, for instance, the difference in coming from the least repressive and most repressive countries in our dataset on the Department of State measure of human rights repression. Similarly, the predicted probabilities for the material and security interests variables, as well as the human rights variables, must be interpreted carefully given that they are interacted with asylum liberalism, and therefore these variables represent effects occurring when asylum liberalism is equal to 0 (essentially the most conservative IJ in our sample). In general, each of these predictors has the expected effect on the granting of relief. As the amount of trade between the United States and the sending country increases, the likelihood of an applicant receiving relief declines. A similar story exists for military aid, although this effect appears to be time bound, having only a modest effect in the full sample (first column) and no effect in models using data from 1999 onward. An applicant from one of the countries historically in the top ten for producing illegal immigrants will have a particularly difficult time attaining relief. Focusing on the full sample, coming from one of these top ten illegal-producing countries decreases the chances of relief by 19 percentage points. An applicant from a more developed country will have a somewhat easier time obtaining relief, as the model for the full sample suggests that an IJ is 6 percentage points more likely to grant relief to someone coming from the most developed countries on this scale as opposed to the least developed countries. Human rights abuse in a country of origin is also predictably important in determining the decision of an IJ. Moving from the least democratic to the most democratic regime decreases the likelihood of relief by between 12 and 18 percentage points, an effect that seems to increase over time, as its effects are larger in the restricted random intercept and stereotype logit models. Our direct measure of human rights abuse within a country is also an important predictor of IJ decision making—coming from a country ranked as one of the worst perpetrators of human rights abuse (as compared to a country respective of human rights) increases the chances of relief by between 20 and 25 percentage points. In the interest of a less muddled discussion of the interaction terms, we hold off discussion of them until we can

Cognitive Approach

71

illustrate effects across the range of asylum liberalism as we do in the various tables and figures referenced in Table 3.1. Looking at the control variables, the results are more or less as we predicted a priori. As we expected, female IJs are 5 to 8 percentage points more likely to vote for some form of relief than are their male counterparts. The language spoken by the applicant also has a small but statistically significant effect on the likelihood of being granted relief, with English speakers and Arabic speakers both about 2 or 3 percentage points more likely to receive relief than applicants speaking other languages. Also important is whether an applicant is represented by legal counsel: we estimate that being represented increases the chances of some form of relief by between 5 and 6 percentage points. To be sure, this is an important effect, but it is not as significant substantively as others have suggested (e.g., Ramji-Nogales, Schoenholtz, and Schrag 2007). This may be because the quality of representation in these cases, in the aggregate, is potentially quite uneven as we discussed above. We hope to explore the role of representation at the IJ stage of decision making more fully in other work, and we revisit this issue when we turn our attention to the decision to appeal an IJ’s decision to the BIA in Chapter 5. When applicants proceed affirmatively with their application, they are more likely to receive some form of relief, between 3 and 5 percentage points more likely, with the effect appearing to increase slightly over time. Detention status has a large effect on the likelihood of relief—those who are detained at the time of their hearing are 12 to 15 percentage points less likely to receive relief than are those who have never been detained. This effect is important given the fact that the number of asylees in detention has increased over time and that many asylees are detained in contravention of the U.S. government’s international commitments (e.g., Human Rights First 2007). There is also some evidence for the effects of national economic conditions on IJ decision making. As the national unemployment rate increases from about 4 percent (5th percentile in the data) to 9.4 percent (95th percentile in the data), the chance of a grant of relief decreases by 8 percentage points, although interestingly we observe this effect only in the data from 1999 onward. This hints at the potential for economic conditions to significantly affect the granting of asylum, and we explore these potential effects more fully in Chapter 4. We have also included a host of control variables for various policy interventions and the September 11 attacks. First, the variable mea sur ing the partisanship of the White House, and therefore the attorney general, is

72

Chapter 3

negatively signed for the random intercept model using all of the data, but is positively signed for the models covering the latter half of our data. We expected that control of the White House by Democrats would increase the rate at which IJs grant relief, but this does not appear to be universally the case. The changing influence of the Democratic administration variable suggests to us that asylum has become an increasingly important policy tool, especially since Nine Eleven. We will further explore the influence of the attorney general in Chapter 5 in the context of the BIA and as part of a strategic approach to understanding IJ decision making. Of note, both of the IIRIRA interventions increased the likelihood of a grant of relief by between 2 and 4 percentage points (these variables are excluded from the models estimated on data from 1999 to 2010) and Real ID also increased the likelihood of relief by 4 percentage points. The overriding impression from the effects of these three variables is that anything that makes it more difficult to move through the asylum process increases the likelihood of those who make it to the IJ stage of receiving relief. The combined effect of these interventions is to raise the likelihood of relief for an applicant by 15 percentage points, a substantial increase that deserves further treatment. We explore the effects of these policy interventions more explicitly, and in a time series framework, in Chapter 6. There is no effect for Nine Eleven—in other words, our model does not find any temporary effect for the attack itself. It remains possible that the attack had a delayed effect on asylum decision making, and we explicitly test whether this is so in our time series analysis in Chapter 6 (and find that there is no delayed effect either, once one accounts for the possibility that Real ID was motivated, at least partially, as a response to the terrorist attacks). Finally, the elapsed time variable is positive and powerful, as we expect. There is a strong trend in our data of increasing grant rates, and moving from the early years in our data to the final years increases the chances of relief quite substantially. One final issue worth exploring before analyzing the interaction terms is determining what, exactly, using the stereotype logit model contributes to our understanding of the asylum decision making of IJs. Looking at the predicted probabilities for the various categories and comparing them to the results for the random intercept model also estimated on the restricted set of data suggests little difference between the results. On the one hand, this is comforting and highlights the overall stability of our approach. But the predicted probabilities displayed in the table are population averages and are not, ultimately, as helpful as they could be in exploring how including a set

Cognitive Approach

73

of ordered outcomes helps us explain decision making. To do so, we turn to simulation and look at somewhat extreme cases. As we discussed in Chapter 2, when we first explored whether there was a logical ordering of outcomes, we view the two intermediate withholding categories as lesser included forms of relief to asylum. Therefore, proceeding with this premise, we should observe the use of these lesser included categories more in cases where the applicant is otherwise less likely to receive asylum and where there are stark differences in the underlying proclivity of the IJ to grant asylum. If we look at a hypothetical applicant who would be otherwise unlikely to receive asylum and compare such an applicant to another who is likely to receive asylum, across the liberalism of IJs, we can begin to see the importance of these lesser included categories. We simulate the probability of someone unlikely to get asylum using the following conditions: in detention at the time of the hearing, fleeing a country with a 0 on the personal integrity scale, and unrepresented by an attorney. We deem someone as relatively likely to receive asylum as not in detention at the time of the hearing, fleeing an abusive country (a score of 4 on the personal integrity scale), and represented by an attorney. Looked at in this manner, we expect that the difference in the propensity of a conservative IJ and a liberal IJ to grant withholding will be greater when the applicant is otherwise unlikely to receive any form of relief than when the applicant is likely to receive asylum. This is exactly what we find: liberal IJs are 73 percent more likely to grant withholding than conservative IJs when other relief is unlikely, but are just 13 percent more likely to grant relief when asylum is otherwise likely. To be clear, the likelihood of being granted withholding remains small whether the IJ is liberal or conservative, but there are real differences in the likelihood of being granted withholding depending on the applicant’s probability of relief and the proclivities of the IJ. Put differently, liberal IJs seem more likely than conservative IJs to rely on the granting of withholding in situations where the applicant would statistically face strong odds of being returned home but are much less likely to do so when the conditions are such that there is a strong possibility of asylum. The evidence suggests that the underlying propensity to grant asylum may influence liberal IJs to be policy innovators and grant a midlevel of relief even though legally withholding of removal requires a higher probability of persecution than does a grant of asylum under U.S. asylum jurisprudence. The implications of this finding are important in terms of international legal standards, in that the finding suggests

74

Chapter 3

that liberal IJs may be more willing to ignore the tougher standard that U.S. jurisprudence has imposed on withholding of removal, and instead they may act more in line with the international norm in regard to the Refugee Convention.

Results: Interaction Terms To economize on space and repetition, we present the results of the interactions only for the model using all of our data in the fi rst column of Table 3.1. The effects of the interaction terms are comparable in the other models, with the exception that later in the data all of the intercepts are shifted higher. Put differently, the relationships between the slopes for the various interactions remain the same, but all judges, be they liberal or conservative, become more likely to grant relief latter in the data. This is concomitant with the generally upward trend in grant rates we have identified and that we explore more fully in Chapter 6. Table 3.2 presents the interaction effects for the two dichotomous variables, military aid and top ten illegal, using a very conservative IJ (5th percentile of asylum liberalism) and a very liberal IJ (95th percentile of asylum liberalism) and various categories of the other relevant variables. In the table we see that across the range of ideology military aid decreases the likelihood of a grant by between 1 and 2 percentage points, with the greater effect occurring for more liberal IJs. Similarly, across the range of ideology for an applicant who comes from a country that is among the top ten producers of illegal immigrants, the suppressing effects are more or less consistent, with a slightly larger effect for liberal IJs than for conservative IJs. Figures 3.1 and 3.2 present the interaction effects graphically for the conservative and the liberal IJ across the range of bilateral trade (Figure 3.1) and World Bank income classifications (Figure 3.2). In general, we get our first hints of the confirmation of our cognitive approach to IJ decision making. The slopes of the lines for the conservative and liberal IJs in Figures 3.1 and 3.2 clearly are not parallel, indicating some degree of contingent evaluation of these material and security interest indicators. For instance, at high levels of trade liberal and conservative IJs become quite similar in their propensity to deny relief. At low levels of development, as indicated by the World Bank classification for a country, liberal and conservative IJs are rela-

Cognitive Approach

75

Table 3.2. Dichotomous Interaction Terms IJ Asylum Liberalism Interaction Term

Conservative

Liberal

Military aid × asylum liberalism Military aid No military aid

0.25 [0.23, 0.28] 0.27 [0.25, 0.29

0.38 [0.31, 0.44] 0.39 [0.33, 0.45]

Top ten illegal × asylum liberalism Top ten illegal Not top ten illegal

0.14 [0.12, 0.15] 0.30 [0.27, 0.32]

0.23 [0.17, 0.28] 0.41 [0.34, 0.48]

tively similar in their decision making as compared to situation in which the applicant is fleeing a well-developed country—in which case liberal IJs are about twice as likely to grant relief as are conservative IJs. Figure 3.3 shows similar changes across the level of democracy in the fleeing country, with strongly parallel slopes suggesting little interpretation of the nature of the sending country’s regime through asylum liberalism. Figure 3.4 displays effects that are also largely noninteractive in nature: the slope of the lines for both liberal and conservative IJs is roughly similar, although a bit steeper for liberals. Finally, note that in all of the figures the intercepts for liberal IJs are higher than those for conservative IJs. This underscores the attitudinal component of asylum decision making: liberal IJs are more likely to grant relief. A more comprehensive understanding of the cognitive hypothesis that we have advanced is possible if we look not at individual variables and their interactions with asylum liberalism in isolation but rather by examining interesting combinations of variables that present a theoretically meaningful profile. We refer to this as a scenario analysis, and the results are presented in Table 3.3. We create two sets of profiles and analyze how the predicted probability of a grant of relief (in the random intercept models) or asylum (in the stereotype model) changes between them across values of IJ liberalism. One set of profiles concerns variation in the material and security interests invoked by the applicant while holding other factors constant. In the high material and high security concern scenario, we set military assistance to 1, top ten illegal to 1, World Bank classification to 0, and log of trade to its 95th percentile value. In the low material and security concern scenario military assistance is set to 0, top ten illegal to 0, World Bank classification to 3,

.6

Predicted Probability of Relief .3 .4 .5 .2 2

4

6

8 Log of Bilateral Trade

10

Conservative IJ

95% CI

Liberal IJ

95% CI

12

.2

Predicted Probability of Relief .3 .5 .4

.6

Figure 3.1. Effect of Change in Bilateral Trade for Conservative and Liberal IJs.

0

1

2

3

World Bank Classification Conservative IJ

95% CI

Liberal IJ

95% CI

Figure 3.2. Effect of Change in World Bank Classification for Conservative and Liberal IJs.

.5 .4 .3

Predicted Probability of Relief

.2 0

1 Level of Democracy (Polity)

2

Conservative IJ

95% CI

Liberal IJ

95% CI

.1

.2

Predicted Probability of Relief .3 .4 .5

.6

Figure 3.3. Effect of Change in Level of Democracy for Conservative and Liberal IJs.

1

2

3 Human Rights Abuse (U.S. State Dept.)

4

Conservative IJ

95% CI

Liberal IJ

95% CI

5

Figure 3.4. Effect of Change in Political Terror Score for Conservative and Liberal IJs.

78

Chapter 3

and log of trade to its 5th percentile value. We then estimate the probability of a vote for relief (or asylum) for a conservative IJ (liberalism set to 5th percentile) and for a liberal IJ (95th percentile of liberalism). We take a similar approach with respect to the human rights concerns raised by an application as well. Specifically, in the high human rights concern scenario we set the personal integrity abuse score to 5 and the democracy variable to 0. In the low human rights concern scenario we set the personal integrity abuse score to 1 and the democracy variable to 2. We are interested in, theoretically, the absolute differences for conservative and liberal IJs across the material and security scenarios and the human rights scenarios, respectively. Our cognitive theory predicts that not all sets of facts will be equally subject to ideological interpretation. The human rights scenarios should constrain IJs more given their official status in the asylum decision-making process. Therefore, we expect the absolute differences between the high and low scenarios to be greater when the facts altered are material and security indicators than when they are indicators of the human rights conditions within a country. We observe, in Table 3.3, precisely what we expected to observe: in each of the models the absolute differences across the material and security scenarios are larger than they are across the human rights scenarios (i.e., the top set of absolute differences in each model is larger than the bottom set). This is evidence that human rights concerns are interpreted less through the lens of asylum liberalism than are the material and security concerns. But note also that there is still a good deal of variation across the human rights scenarios, which is evidence that attitudes still matter even when interpreting facts that are legally relevant in the U.S. asylum context. The differences in the magnitude of the absolute differences range from 21 percent in the stereotype logit model for a liberal IJ to 65 percent for a liberal IJ in the random intercept model with all of the data. Furthermore, within each of the scenarios the relative ordering of the likelihood of granting relief (or asylum) is as we would expect: conservative IJs faced with high material and security concerns are least likely to grant (between 9 and 11 percent) and liberal IJs faced with low material and security concerns are most likely to grant (between 69 and 82 percent). Interestingly, then, another takeaway from this analysis is that it is more the consideration of extralegal factors than of legal factors that drives the variation in IJ grant rates so lamented by law professors and commentators. Th is implicates, once again, norms of fairness and consistency in terms of both domestic and international law.

Table 3.3. Scenario Analysis Model 1: Random Intercept Logit (1990–2010) Conservative Policy Preferences High material & security interest Low material & security interest High human rights concerns Low human rights concerns

Liberal Policy Preferences

0.09 [0.08, 0.11]

0.14 [0.11, 0.18]

0.57 [0.54, 0.59]

0.79 [0.75, 0.84]

0.46 [0.44, 0.49]

0.61 [0.55, 0.67]

0.15 [0.13, 0.16]

0.24 [0.19, 0.29]

Conservative Absolute Difference

Liberal Absolute Difference

0.48

0.65

0.31

0.37

Model 2: Random Intercept Logit (1999–2010) Conservative Policy Preferences High material & security interest Low material & security interest High human rights concerns Low human rights concerns

Liberal Policy Preferences

0.11 [0.09, 0.12]

0.21 [0.16, 0.27]

0.59 [0.56, 0.62]

0.82 [0.77, 0.87]

0.53 [0.50, 0.56]

0.66 [0.60, 0.73]

0.17 [0.15, 0.19]

0.37 [0.30, 0.44]

Conservative Absolute Difference

Liberal Absolute Difference

0.48

0.61

0.36

0.29

Model 3: Stereotype Logit (1999–2010) Conservative Policy Preferences High material & security interest Low material & security interest High human rights concerns Low human rights concerns

Liberal Policy Preferences

0.11 [0.08, 0.14]

0.20 [0.10, 0.29]

0.51 [0.44, 0.57]

0.69 [0.55, 0.82]

0.50 [0.44, 0.54]

0.58 [0.47, 0.69]

0.19 [0.15, 0.22]

0.37 [0.30, 0.45]

Conservative Absolute Difference

Liberal Absolute Difference

0.40

0.49

0.31

0.21

80

Chapter 3

Finally, we return to the IR literature and Rosenblum and Salehyan’s (2004) work to inform this profi le. As we noted above the most theoretically interesting quadrants from the norm versus interest perspective in their two-dimensional analysis identified were those categories in which humanitarian concerns and interests would predict contradictory outcomes: normative interests predict grant but U.S. interests predict denial or U.S. interests support a grant but normative interests are lacking. The above scenarios do not capture this conflict, so we conduct one more analysis with predicted probabilities for liberals and conservatives when human concerns are at their highest level and thus typically would predict a grant and national interests are also at their highest and would typically predict a denial. Table 3.4 provides the results of this analysis. Here we find that conservatives are 14 percentage points less likely than liberals to grant asylum. This difference is larger than the ideological difference we saw in Table 11 when human rights concerns are at their highest but national interests are held to the mean (a negative 10 percentage point difference) and is higher than when national interests to deny are at their highest with human rights concerns held to the mean (a negative 8 percentage point difference). Thus when conflict between the two competing sets of values is greatest, the ideological differences are greatest as well. In terms of the cognitive theory our expectations hold here as well. When we move from a scenario in which national interest to deny is highest with human rights only average to a scenario in which there is direct conflict between high human rights interest to grant and high national inter-

Table 3.4. Human Rights and Material and Security Concerns Contrasted, 1999–2010 Model 3: Stereotype Logit (1999–2010)

Scenario High material & security interest; high human rights concern Low material & security interest; low human rights concern

Conservative Policy Preferences

Liberal Policy Preferences

0.18 [0.14, 0.22]

0.32 [0.18, 0.45]

0.17 [0.10, 0.24]

0.30 [0.18, 0.42]

Conservative Absolute Difference

Liberal Absolute Difference

0.01

0.02

Cognitive Approach

81

est to deny, IJs’ probability to grant decreases by 7 percentage points for conservative IJs and 12 percentage points for liberal IJs. To return to our cognitive and attitudinal theory, in each of the models the absolute differences across the material and security scenarios are larger than they are across the human rights scenarios. This is evidence that human rights concerns are interpreted less through the lens of asylum liberalism than are the material and security concerns.

Conclusion The analyses in this chapter have theoretical implications and encompass several salient policy issues. In terms of theory, the analyses presented here expand and improve upon the compliance literature by examining actual state actors on the ground who engage in behavior controlled by U.S. international commitments. We demonstrate that while evidence linked to international refugee norms clearly matters in the asylum decisions, the dominant factor that influences a grant of asylum is a matter of “the luck of the draw” so to speak. What matters most is the IJ whom the applicant faces. In this chapter we take the important step of integrating theories from IR and judicial behavior, applying a version of the attitudinal model. We specify an underlying causal mechanism for the operation of these preferences in asylum cases: a cognitive process that accounts for the individual weight that a judge may give national security, economic issues, or human rights cues, while controlling for a variety of factors that may influence the judges’ decisions. We demonstrate that the large effects for immigration liberalism primarily enter the asylum decision-making process through the extralegal consideration of the material and security interests of the United States whereas humanitarian concerns impose some constraint on the IJs’ use of policy preferences. While previous studies in the IR literature have shown the role of both national interests and humanitarian concerns, we are able to offer a cognitive model that explains how these factors play out in the IJs’ decision-making process. Thus, the underlying implication for political science is the importance of crossdiscipline research, in the sense that IR approaches benefit from the inclusion of theories of judicial behavior by specifying an underlying causal mechanism for many of the central findings in relation to “state” action. In regard to the work of law professors, the findings in this chapter strongly emphasize the need for theory to understand data. Given the application of a

82

Chapter 3

blended cognitive/attitudinal theory of judicial decision making to IJs, we are, for the first time to our knowledge, able to specify how variation enters into IJ decision making in a manner that has previously been unexamined: IJs feel somewhat constrained with respect to legally relevant facts (human rights conditions), but still evaluate applicants using a process that is driven by their policy proclivities. Th is decision-making approach is enabled by high workloads, relatively little solid evidence with respect to persecution, and, as we will demonstrate in subsequent chapters, a great deal of variation in the support available to litigants locally and fairly predictable review from the BIA. Our analysis also triggers policy concerns about the fairness and consistency of asylum decisions because a grant or denial is predicated on the IJs’ policy preferences. The bad news is offset by the demonstration that the law does matter and its effects appear to be less influenced by the IJs’ political preferences. But then again, other, extralegal, factors influence the IJs’ decisions through these policy proclivities. It is beyond this book to take on the entire reform issue surrounding immigration courts; however, one recommendation is clear. The variation in the treatment of seemingly like cases should be limited by reducing the reliance on factors that seem to be driven by U.S. strategic, as opposed to humanitarian, considerations. Such reforms would likely include providing more job security for IJs and more rigorous oversight from the BIA, in addition to more cultural and legal training for IJs. Whether or not reducing the caseload and increasing IJ support would increase consistency is not known, because while allowing IJs more time to decide each case will undoubtedly mean less reliance on heuristics and more deliberative and reasoned decision making, it will not necessarily reduce variation in the decisions reached because it may just allow more room for IJ preferences to enter into the process. Furthermore, it is possible that increasing the amount of time the IJs have to consider each case could increase the influence of extralegal factors. Our findings of the influence of career socialization suggest that there should be greater consideration of this reality and its implications, especially in regard to normative concerns and in regard to the numerous reforms that have been proposed. As scholars of the Supreme Court have long known, attitudinal preferences are like water—they find a way into the decisionmaking process. Considering how those preferences enter into decision making is crucial to understanding the plausible ways inconsistency between decision makers might be reduced. We return to reform suggestions

Cognitive Approach

83

in Chapter 7 when we can discuss them in the context of our fuller set of findings from across all chapters. Of course, our conclusions are limited by inherent data limitations: applicant and case data are limited by congressional mandate. Understanding of the adjudication of asylum claims would be greatly enhanced if Congress softened the privacy mandate they have given to EOIR. To date, no nongovernmental entity has been granted access to these types of core data, such as the types of evidence presented by asylum seekers. While in the analyses in these chapters we give considerable attention to IJs and their decision making, we are not able to control for or test fully the influence of location on the IJs’ decisions. In the next chapter we develop and test a local and national political economy framework of IJ decision making.

CHAPTER 4

Local Conditions and IJ Decision Making

In this chapter we expand the cognitive model we presented in Chapter 3 and in Keith, Holmes, and Miller (2013) to integrate expectations derived from the long-standing debates within the immigration literature to further understand the asylum decisions of immigration judges (IJs). It is important to understand how local conditions might affect IJ decision making. If local factors influence who gets asylum, then the prospects for reducing the variability between adjudicators are somewhat dimmed. Within the asylum literature there is scant work examining whether local economic or demographic conditions have an impact on asylum decisions, unlike the broader immigration literature, which focuses on the competing expectations of the role of threat and contact to explain anti-immigrant sentiments, approaches that are very much focused on the local context. In this literature the primary debate centers on whether increased interaction with immigrants can either incite hostility or promote acceptance. Here we seek to integrate those approaches with our models of asylum decision making. The threat hypothesis posits that threats to cultural identity or economic threats drive perceptions about immigrants, whereas the contact hypothesis posits that meaningful contact with immigrants promotes tolerance (see Cornelius and Rosenblum 2005 and citations within; see also Fetzer 2000). In terms of conflict, unsuccessful integration can increase cultural clash (see Ireland 2004) or instigate a perception of cultural threat (Rudolph 2003). Immigrants may also be perceived as an economic burden on the receiving community. Specifically, we examine two alternative approaches in regard to economic conditions and perceptions of threats. First, Nicholson-Crotty and Nicholson-Crotty (2011) distinguish among industrial sectors to examine the effect of local industry labor demand on immigration policy. They argue that areas with core industries that rely heavily on undocumented la-

Local Conditions

85

bor will have policies more supportive of immigrants. Second, Hopkins’s (2010) “politicized places” hypothesis posits a conditional relationship between broad concern about immigrants and how many residents are foreignborn. Specifically, when there is a rapid increase in the foreign-born population and the issue of immigration is highly salient, there will be a backlash against immigrants in the local community. IJs are not well understood, as they are not typical judges, as discussed in previous chapters. It is possible that they may be more sensitive to economic and demographic pressures within their local communities given their relative lack of job security and the ease with which they can be identified with particular policy choices. In this chapter, we examine metropolitan areas (metropolitan statistical areas or MSAs) to explore if judges are responsive to local effects. In our analysis we have IJs who sit on the bench in thirty-two different metropolitan areas and make decisions in approximately 129,000 cases over the time span of our data. Although there are fi ft y-five immigration courts, sometimes there are multiple courts in the same metropolitan area. For example, in California, both the Los Angeles and Lancaster courts are in the Los Angeles metropolitan area, with similar clusters of immigration courts in New York/Newark, Phoenix, Washington, D.C., and Miami. The MSA is the smallest meaningful geographic unit that has reported economic and demographic statistics. In the sections that follow, we integrate the theoretical approaches discussed above regarding contact, threat (cultural and economic), and politicized places into our previous model and empirically test our expectations by examining all asylum decisions by IJs from 2005 to 2010.1 In our analyses we find evidence that IJs are responsive to their local communities, sometimes in ways that are conditional upon the judges’ ideology. Furthermore, we find some support for each of the approaches that other scholars have focused on.

Expanding the Cognitive Model There is evidence from studies of the behavior of federal and state judges that decision making can be influenced by local conditions. For example, it has been demonstrated that federal district court judges, insulated though they are from political pressures by the granting of life tenure, are susceptible to local conditions in their decision-making processes. Studies of southern district court judges asked to enforce desegregation orders have shown that patterns of

86

Chapter 4

enforcement tend to vary with local sentiment on the desirability of desegregation (Vines 1964; Peltason 1971; Giles and Walker 1975). Similar results have emerged from studies of the attempt of state supreme court judges to enforce controversial Supreme Court decisions on the constitutionality of prayer in school (Tarr 1977), the decisions of federal district court with respect to the prosecution of Vietnam War protestors’ convictions (Kritzer 1978), and decision making of federal circuit court judges in environmental cases (Wenner and Dutter 1988). IJs have tenure protections that are significantly weaker than all federal Article III judges and some state judges. Therefore, given the evidence of responsiveness for these more insulated judges, we have reason to believe that less protected IJs might also be likely to respond to local conditions. We argue that in the context of IJ decision making a very particular set of local conditions is likely to influence the IJ’s decisions. We derive our understanding from the broader immigration literature, which we discuss more fully below. We believe that these extralegal factors, similar to those related to material and security concerns, are likely to be evaluated in a more top-down fashion (or more subjectively) than legal factors and thus will be contingent upon the IJ’s policy predispositions. This expectation represents our underlying general hypothesis—that IJs will be affected by local conditions but that those conditions might matter differently to different IJs. That is, similar to our treatment of applicant characteristics in Chapter 3, we believe that IJs might weigh these extralegal factors differently, depending on their own ideological proclivities. In this manner, our investigation here represents an extension of the cognitive model we first explored in Chapter 3, but here we focus on how relevant local conditions might influence IJ decision making, and particularly how those conditions interact with (or are weighted by) the ideology of the IJs. If IJs not only react to different local conditions, but also react differently to the same local conditions, then significantly reducing the variability in their decision making may be more difficult than first contemplated. We turn next to the literature on contact and threat theory.

Contact Theory

Contact theory has somewhat divergent expectations on whether the presence of immigrants within a community has the effect of encouraging tolerance or support for immigrants or the effect of increasing tension (Holzer, Schneider, and Widmer 2000, 263). As applied to asylum decision making, the per-

Local Conditions

87

centage of the population that is foreign-born or the presence of immigrant communities can be theorized to have a positive impact on asylum seekers because there is a welcoming community for immigrants. Yarnold (1994, 210) argues that “where the concentration of immigrants in a judge’s area is high, the judge may be more inclined to rule in favor of a refugee than a judge in a low-immigrant-flow area, since immigrants in a high-flow area may, for example, form groups which constitute a pro-immigrant lobby.” Similarly Boswell (2003, 324) states, “[S]ocial tension is usually highest in areas with relatively small numbers of asylum seekers and little experience of integrating other ethnic groups.” In addition, some studies of public attitudes toward undocumented workers suggest that increased contact between immigrants and the native population will decrease tensions between them (Hood and Morris 1998). Alternatively, a high proportion of foreign-born can also have a negative impact, as prejudice against immigrants can increase with more contact. Holzer, Schneider, and Widmer (2000, 261) examined asylum decision making in Switzerland and argued that the effect of the proportion of foreign residents within a canton could work in two directions: “Obviously, a large share of foreigners makes contacts between immigrants and the holder of a Swiss passport more likely. This can either diminish negative prejudices against foreigners in general and asylum seekers in particular or, in the case of negative experiences, enhance them.” Therefore, there is support in the literature for contradictory expectations that contact with foreigners will increase or decrease tolerance for them. Translated to IJs this means that the presence of large proportions of foreign-born residents in a metropolitan area may influence IJs to increase the rate at which they grant relief to applicants or to decrease that rate. There is also a voluminous literature on what has come to be called “chain immigration,” a phenomenon in which a particular area becomes a popular destination for specific ethnic groups. Such immigration is particularly likely to occur in the United States because of immigration policies that favor family reunification (e.g., Jasso and Rosenzweig 1989) and because immigrants are attracted to areas in which there are likely to be support networks accessible to them (e.g., Koser 1997). We conceptualize these as diaspora communities and hypothesize that these diaspora communities might constitute an additional contextual (contact) advantage to asylum seekers or might alternatively operate as a more specific threat. IJs might be more liberal in their granting of asylum when asylum seekers apply in areas with large numbers of co-ethnics because they may believe that these applicants are more likely to be supported and

88

Chapter 4

integrated into the community. In this sense, the explanation for a positive effect for a large diaspora community may be as much economic as cultural. On the other hand, IJs might perceive an increased threat from those migrating to already large communities. Or, and probably more likely in the context of an asylum application in which credibility is often the central issue, IJs might perceive those migrating to a diaspora community as less credible than other applicants, simply following on in the path of previous family and friends, pursuing economic opportunities and not genuinely fleeing persecution.

Economic Threat

Asylum seekers themselves are thought to be sensitive to the economic conditions in both their country of origin and the destination country, in particular unemployment and the perceived overburden of asylum seekers in the country of destination (Neumayer 2005, 64). It is not surprising that there is a growing literature that seeks to determine how best to share the fiscal burden of receiving refugees and asylum recipients across states or other subunits of government. For example, Boswell (2003, 317) examines the British and German efforts to disperse asylum recipients across regions within their countries in response to the increasing costs and public scrutiny of asylum recipients. This effort was initiated in response to economic crisis and increasing unemployment, coupled with the growing resentment toward the “perceived generosity of the social benefits offered to asylum seekers.” As Cornelius and Rosenblum (2005, 103) note in regard to the U.S. context, “immigrants are likely to represent a net fiscal burden to the localities in which they settle” and “there is no question that immigrants to the United States are a fiscal drain on states and localities with large immigrant communities,” despite being “net contributors to the federal treasury.” Research has also demonstrated that public opposition to immigration increases in response to a vulnerable labor market, as delineated by such factors as unemployment and anxiety about one’s job security (Citrin et al. 1997; Alvarez and Butterfield 2000). This finding parallels research on public hostility to immigrants in which unemployment itself as well as the salience and perception of unemployment as a problem was associated with more hostility toward immigrants (Hoskin 1991). Interestingly, in two U.S. studies of asylum-related appeals in federal district and circuit courts from 1980 to 1997, neither unemployment nor regional effects where significant, when controlling for appointing president and high

Local Conditions

89

immigrant flow (Yarnold 1990, 1994). We wish to emphasize that these cases were appeals of prior immigration decisions made to the Board of Immigration Appeals, and thus these are decisions that are several steps removed from the choices made by IJs themselves and are decisions by other actors with different institutional norms and constraints. Ultimately, we believe that IJs will decrease their grant rates in light of deteriorating local labor markets, although it is also plausible that IJs will see applicants who continue to pursue a claim in the face of a deteriorating labor market as more credible than they otherwise would (or at least less likely to be an opportunistic asylum seeker).

Immigrant-Friendly Industries

Nicholson-Crotty and Nicholson-Crotty (2011, 612) move beyond a simple conceptualization of the effect of labor market on the perception of immigrants to focus on industries that rely heavily on “immigrant labor for their profitability and survival.” Although their focus is on immigrant policies, their work supports the conclusion that “states where these industries are more politically powerful should have less restrictive immigrant policies as these groups pressure lawmakers to pass policies that make the state attractive to the immigrant labor they require” (614). Similarly, we expect that metropolitan areas with higher proportions of industries that rely upon illegal immigrant labor may be more amenable to immigrants in general, providing subtle pressure on judges to be more lenient. Metropolitan areas with specific industries that depend upon “immigrant labor for their profitability and survival” such as agriculture and textiles may have more pressure for less restrictive immigration policies (Nicholson-Crotty and Nicholson-Crotty 2011, 612). Therefore, in places where immigrant-based industries are more prominent IJs might be more willing to grant relief to applicants, believing them more likely to find gainful employment.

Politicized Places

We also consider Hopkins’s (2010) politicized places hypothesis that posits an alternative theory in regard to how local and national conditions interact to influence perceptions of immigrants as threats. He notes that despite the

90

Chapter 4

strong expectation that immigrant-heavy communities will have more anti-immigrant attitudes, the empirical evidence has been inconclusive. He suggests instead that “sudden demographic changes generate uncertainty and attention” and that the media inform the public about these changes and that can “politicize those changes in people’s minds” (40). Hopkins argues that “people are highly selective in incorporating environmental information” and that “when people fi lter the vast quantities of information available, they pay special attention to change” (42 and citations within). The public may not immediately connect those changes to political or economic ramifications, but media attention and rhetoric provide a frame that identifies and clarifies the ramifications. Thus Hopkins hypothesizes that “hostile political reactions to neighboring immigrants are most likely when communities undergo sudden influxes of immigrants and when salient national rhetoric reinforces the threat” (41–42). The framing component of Hopkins’s argument may not be relevant for IJs, since immigration is constantly a salient issue—every case before IJs involves immigrants, and they deal with the local flow of immigrants on a daily basis. Although Hopkins’s hypothesis is clearly meant to apply to the public, we argue that there are two routes by which the politicization of immigration and backlash in a particular place might affect IJs. The less likely route is direct in the sense that IJs might also become more hostile to asylum seekers as the public in a metropolitan area experience backlash—this would constitute a direct influence on the opinion of IJs toward immigrants. The second, more likely route for an effect for politicized places is that IJs fear a backlash against their decisions to admit asylees rather than being directly affected by a community’s backlash against immigration, something that scholars of judicial behavior might call a strategic reaction to the potential for backlash from the community. Unfortunately, our observational data do not allow us to distinguish between these two casual mechanisms. Nevertheless, we believe the indirect (or strategic) route of influence makes more sense than a direct route because we think it is unlikely that IJs themselves, dealing everyday with large numbers of immigrants, are likely to change their opinions with the ebbs and flows of local attention to the issue. To summarize, we have the following hypotheses for each of the three major theories. Contact/Threat: Hypothesis 1: As the proportion of a population increases in an MSA, IJs will become more likely to grant relief (contact theory).

Local Conditions

91

Hypothesis 2: As the proportion of a population that is foreign-born increases in an MSA, IJs will become less likely to grant relief (cultural threat). Hypothesis 3: If the applicant is migrating to a community with a large matching diaspora community, the IJ will be more likely to grant relief (contact theory). Hypothesis 4: If the applicant is migrating to a community with a large matching diaspora community, the IJ will be less likely to grant relief (threat theory). Economic Threat: Hypothesis 5: As the unemployment rate in a local community increases, an IJ will be less likely to grant relief. Hypothesis 6: As the ratio of immigrant-heavy industries in an MSA increases, an IJ will be more likely to grant relief. Politicized Places: Hypothesis 7: When there is rapid change in the percentage of the population that is foreign-born and immigration is nationally salient, IJs should be less likely to grant relief. These hypotheses represent unconditional expectations. However, congruent with our cognitive approach to IJs we believe that conservative and liberal IJs are likely to react differently to some, if not all, of these conditions, hence our inclusion of interactions. We do not have specific a priori expectations with respect to these conditional effects, and so we do not present hypotheses with respect to them. Nevertheless, such differing reactions would be particularly helpful in understanding whether contact with foreigners or cultural threat from them is a conditional explanation depending on the ideology of the actor, given the divergent expectations for the presence of large numbers of foreigners.

Describing the Local Contexts of Asylum Decision Making Before discussing the data used in the models that follow, we pause to discuss the MSAs we include in our analyses. Table 4.1 lists some pertinent characteristics of the included metropolitan areas, such as the number of included

Table 4.1. Descriptive Statistics for Included MSAs, 2005–2010

MSA Atlanta Baltimore Boston Buffalo Charlotte Chicago Cleveland Dallas Denver Detroit El Paso Hartford Houston Kansas City Las Vegas Los Angeles Memphis Miami Minneapolis

Observations

Percentage of Total Cases

1,338 3,826 4,398 578 97 2,958 2,321 1,016 1,443 2,301 497 1,096 2,612 756 652 13,669 1,919 24,360 1,408

1.03 2.96 3.40 0.45 0.08 2.29 1.80 0.79 1.12 1.78 0.38 0.85 2.02 0.58 0.50 10.60 1.48 18.84 1.09

Mean Relief Rate

Percentage ForeignBorn

Average Unemployment

Average Immigration Employment Ratio

0.15 0.51 0.35 0.19 0.54 0.56 0.44 0.40 0.37 0.30 0.26 0.30 0.16 0.28 0.45 0.43 0.51 0.20 0.35

0.14 0.08 0.16 0.05 0.10 0.18 0.06 0.18 0.12 0.09 0.26 0.12 0.21 0.06 0.22 0.35 0.04 0.37 0.09

6.56 4.91 5.81 6.25 10.67 6.92 7.24 6.04 6.21 10.35 7.49 6.43 5.62 7.64 7.37 7.69 7.14 5.59 4.86

0.062 0.056 0.056 0.068 0.057 0.069 0.059 0.057 0.058 0.056 0.064 0.054 0.061 0.059 0.094 0.072 0.066 0.065 0.052

New Orleans New York Omaha Philadelphia Phoenix Portland Salt Lake City San Antonio San Diego San Francisco Seattle Tucson Washington, D.C. Total/Average

361 36,652 859 2,839 2,004 586 327 1,358 1,914 9,320 2,575 242 2,972 129,254

0.28 28.35 0.66 2.20 1.55 0.45 0.25 1.05 1.48 7.21 1.99 0.19 2.30 100

0.34 0.71 0.23 0.32 0.40 0.29 0.46 0.39 0.58 0.46 0.43 0.74 0.68

0.06 0.28 0.06 0.09 0.17 0.13 0.11 0.11 0.22 0.29 0.15 0.14 0.20

6.32 5.61 4.49 7.07 6.81 7.13 4.84 5.48 6.49 5.69 6.81 5.32 2.86

0.066 0.063 0.056 0.055 0.067 0.058 0.058 0.061 0.066 0.067 0.054 0.059 0.058

0.46

0.17

6.08

0.064

94

Chapter 4

observations from each MSA, the average relief rate within each, the average level of the foreign-born population, average rates of unemployment, and the portion of the labor force that works in immigrant-labor-intensive industries. As noted above, there are thirty-two MSAs included in the analyses, although there are fi ft y-two immigration courts. For some immigration courts we could not collect the relevant data, and so they are excluded from our analyses (for instance the immigration court in San Juan); other immigration courts are subsumed within a single MSA; for instance, the Krome detention facility and the Miami immigration court are both located in the Miami MSA. For the analysis in this chapter, our data include 73 percent of the total asylum decisions made between 2005 and 2010. Table 4.1 reveals that cases are in no way evenly distributed across these metro areas, with the New York and Miami courts, combined, contributing over 46 percent of the cases—indeed the four busiest MSAs contribute 63 percent of the cases in our data (New York, Miami, Los Angeles, and San Francisco). There is tremendous variation across the MSAs with respect to the mean relief rate.2 Atlanta is the most difficult place to gain relief, with just 15 percent of applicants getting any form of relief; meanwhile, New York is the most lenient MSA, as 71 percent of applicants are granted some form of relief. There is also a good deal of variation in the average levels of foreign-born living in these areas. Not surprisingly, Miami, Los Angeles, San Francisco, and New York have the largest foreign-born populations, with significantly lower concentrations of foreigners in smaller cities like Buffalo, Cleveland, Kansas City, Memphis, New Orleans, and Omaha. There is also considerable variability in the economic situations in these metro areas, with unemployment averaging a high of 10.67 percent in Charlotte and a low of 2.86 percent in Washington, D.C. There is also some notable variation in the importance of immigrant-labor-intensive industries in these communities. Las Vegas has an usually high concentration, at almost 9.4 percent of the total labor force, whereas most of the MSAs average around 5.5 percent of the labor force employed in these industries.

Data Description Dependent Variable

Our unit of analysis is the decision of an IJ on the merits. We use the fourpoint coding scale we initially discussed in Chapter 2, with no relief equal to 0,

Local Conditions

95

withholding under CAT equal to 1, withholding equal to 2, and asylum equal to 3. Descriptive data on the dependent variable and the other included variables included in these analyses are included in the online appendix to this chapter.3

Economic Threat

Within the United States, there is significant variation among metropolitan economic conditions that could influence individual IJs. The local unemployment rate from the court’s metropolitan area is our measure of local economic conditions, as provided by the Bureau of Labor Statistics. To create the ratio of immigrant employment to test Nicholson-Crotty and NicholsonCrotty’s expectation that areas with a higher proportion of industries that hire undocumented workers would have more liberal immigration policies, we utilized the Occupational Employment Statistics from the Bureau of Labor Statistics. Here, we use the occupations identified by Passel (2006, 12) as having more than three times the average share of unauthorized workers. The national average is 4.9 percent, so we code industries with a percentage of unauthorized workers that ranges from 15 to 36 percent. The estimate is from March 2005 and derived from the Current Population Survey. Passel (2006) identified the industries using the standard North American Industry Classification System.

Contact Theory

Higher exposure to immigrants can either trigger a threat response or promote tolerance (Cornelius and Rosenblum 2005). To explore these competing expectations, we include a measure of the percentage foreign-born in an MSA by year, using U.S. Census Bureau data. We also include a dummy variable to identify if an asylum seeker is a member of a diaspora community in the metropolitan area.4 These data are coded using maps provided by the Migration Policy Institute based upon the 2008 American Community Survey (ACS) of the U.S. Census Bureau. Significant diaspora communities were identified by the Migration Policy Institute as a foreign-born population that exceeded the following thresholds in metropolitan areas: more than 150,000 Mexicans, 35,000 Filipinos, 35,000 Koreans, 30,000 Indians,

96

Chapter 4

30,000 Chinese, 25,000 Salvadorans, or 30,000 Viet namese, respectively, in each metropolitan area. The Migration Policy Institute identified these thresholds by selecting the top ten metropolitan areas that had concentrations of particular immigrant groups in the 2008 ACS.5

Politicized Places

Finally, to check the conditional relationship as theorized by Hopkins in his politicized places theory, we included his original USA Today measure of national immigration salience, which is simply a count of the number of stories mentioning immigration per month in USA Today. Two things are noteworthy about this measure of salience, both gleaned from Hopkins (2010). First, we use USA Today because it is the most widely read national newspaper in the United States. Second, stories in USA Today on immigration are overwhelmingly negative, and therefore the count of stories should have a negative effect on grant rates. Second, we include the percentage change in the foreign-born population between 2000 and 2005 using U.S. Census Bureau data. The essence of the politicized places hypothesis is that communities that experience a rapid influx of immigrants when immigration is salient will be likely to react negatively to that influx, which is captured by an interaction between the rate of change in the foreign-born population and immigration salience. Our question is different from the one posed by Hopkins in that our actors (IJs) are political elites for whom fluctuations in the salience of immigration may not matter considerably. On the other hand, it seems worth testing whether IJs are responsive to communities that may be experiencing backlash, and so for this reason we include a three-way interaction among change in the foreign-born population, national immigration salience, and IJ liberalism.

IJ Liberalism

We have already demonstrated that the policy predispositions of the IJs play a dominant role in explaining the observed discrepancies in asylum grant rates and that liberal IJs respond to certain applicant characteristics differently than their more conservative colleagues. We use the same factor score of IJ career socialization to create a proxy for a policy predisposition toward

Local Conditions

97

immigration rights and asylum to serve as a proxy for a policy proclivity to grant or deny, as we used in Chapter 3. We include nine control variables, seven of which we discussed previously in Chapter 3, and we do not reiterate that discussion here. There are two additional control variables that do not appear in models in Chapter 3 that we include here. First, we include a count of the number of local recognized and accredited asylum nongovernmental organizations (NGOs) in an MSA.6 The number of NGOs is calculated based on a list provided by the Executive Office for Immigration Review (EOIR). We counted each organization that was within a one-hour drive of a given court. Presumably, the more asylum-related NGOs nearby, the greater the likelihood that an individual will receive specialized support and legal advice. We also include a measure of the local political environment based on the average Democratic vote for president by counties within an MSA in 2004 (for 2005, 2006, and 2007) and 2008 (for 2008, 2009, and 2010). Descriptive statistics for the model as well as diagnostic checks for robustness are included in the online appendix that accompanies this chapter.

Analysis Our data structure is a bit complex. As before, applicants nest within IJs, but now we also have data on MSAs, in which IJs nest. This means that we have a three-level data structure, with applicants at the bottom level, IJs at the midlevel, and MSAs at the top level. A natural approach to modeling these data would be to utilize a three-level hierarchical model. Unfortunately, given the number of observations in our data (over 129,000), the large number of covariates, particularly interaction terms, and the relatively small number of level-three units (there are only thirty-two MSAs included in the data), we could not estimate such a model. Instead, we estimate two separate stereotype logit models and allow the data to cluster either on IJs or on MSAs. Table 4.2 provides estimated effects for a stereotype logit model in which the errors are clustered by IJ. An alternative specification, with errors clustered by MSA instead of IJ, generates similar results. Both models are generally in agreement about the effects for the variables, with similarly signed and sized coefficients, minimal differences in standard errors, and highly similar ancillary parameters. The parameter estimates for the two models are available for comparison in the online appendix. Here, in an attempt to clarify

98

Chapter 4

Metropolitan Statistical Area (MSA)—32 Total

Immigration Judge (IJ)—271 Total

Asylum Applicants— 129,284 Total

Figure 4.1. Nesting Structure for Local Data.

our interpretive task somewhat, we do not carry over interactions from the models in Chapter 3. We do this because there are a large number of interactive terms necessary for testing the conditional impact of local conditions on different types of IJs and additional interactions further muddle interpretive clarity. We have three broad sets of hypotheses (economic threat, contact/cultural threat, and politicized places) which we believe might have effects that are conditional on the ideology of the IJ, thus necessitating a battery of interaction terms. First, many of the control variables are significant, although English speaker has a substantively small effect. The most notable effect for the control variables is for our count of the number of registered NGOs in a community. As this count increases from 8, the number in San Antonio, which is the 10th percentile in the data, to New York’s 88, which is the 90th percentile, the likelihood of a grant of asylum increases by a substantial 23 percentage points. Such an effect is surprising in its magnitude, given that we are already controlling for representation, which also increases the likelihood of asylum by 13 percentage points. This suggests that these NGOs have a powerful effect on the process by doing things other than simply providing representation. It may be that they also have a tight connection with the

Table 4.2. Stereotype Logit Model Probability of Grant of Relief (2005–2010)

Variable Judicial policy preferences Asylum liberalism (+) Applicant characteristics Log of trade with U.S. (−) U.S. military aid (−) Top ten illegal immigration (−) World Bank development class (+) Democracy (Polity) (−) Human rights abuse (PTS-State Dept.) (+) Economic threat MSA unemployment (−) Immigration employment ratio (+) Contact/threat Percentage pop. foreign-born (+/−) Diaspora community (+/−)

Clustered by Judge Change in Pred. Prob. of Asylum see Tables 4.3−4.5 −0.19 [−0.19, −0.19] — −0.20 [−0.20, −0.20] 0.07 [0.06, 0.08] −0.16 [−0.16, −0.15] 0.18 [0.17, 0.19] — — — 0.04 [0.03, 0.05]

Politicized places National immigration salience (~) Change in % foreign-born (~) National immigration salience × change in % foreign-born (−)

0.04 [0.04, 0.04] — −0.01 [−0.03, −0.01]

Interactions MSA unemployment × asylum liberalism Immigration employ. ratio × asylum liberalism Percentage pop. foreign-born × asylum liberalism Diaspora comm. × asylum liberalism Immigration salience × asylum liberalism Change % foreign-born × asylum liberalism Immigration salience × change % foreign-born × asylum lib.

see Table 4.3 see Table 4.3 see Table 4.4 see Table 4.4 see Table 4.5 see Table 4.5 see Table 4.5

Controls Gender (+) Legal representation (+) English speaker (+) Arabic speaker (+/−) Detention status (−) Affirmative application (+) Democratic admin. (+) MSA democratic vote (+) Registered NGOs (+)

— 0.13 [0.12, 0.13] 0.02 [0.02, 0.03] — −0.11 [−0.13, −0.10] 0.08 [0.07, 0.08] 0.07 [0.06, 0.09] — 0.23 [0.21, 0.25]

Note: 95% confidence intervals in brackets; predicted probabilities calculated only for statistically significant coefficients (p < 0.05, two-tailed).

100

Chapter 4

community in terms of ser vices that begin to integrate the applicant into the community, and some of the NGOs may have parallel refugee resettlement programs. NGOs are likely to be able to help otherwise indigent applicants build their cases for credibility by, for instance, connecting them with country experts who may be able to testify to the credibility of the applicant’s testimony. Those in detention are 11 percentage points less likely to receive asylum than those who have never been detained, and those who file affirmatively are 8 percentage points more likely to receive asylum. Finally, when the administration is controlled by Democrats, applicants are about 7 percentage points more likely to receive asylum. This is evidence for some degree of hierarchical control, a topic we explore in more detail in Chapter 5. The effects for the various applicant characteristics, including material and security interests as well as human rights conditions, are substantively important. IJs continue to react very negatively to those applicants arriving from countries that are among the top ten producers of illegal immigrants, as all else equal these applicants are 20 percentage points less likely to garner relief. IJs are also more likely to grant relief to asylum seekers who are fleeing more prosperous economies as measured by the World Bank. Moving from among the least developed countries to the most increases the probability of a grant by 7 percentage points. Similarly, those coming from prominent U.S. trading partners are also less likely to receive relief, by 19 percentage points. Military aid does not have a statistically significant effect on the likelihood of relief. Human rights conditions also continue to figure prominently in the model, with both measures registering as statistically significant. Coming from a democratic regime, versus a nondemocratic regime, decreases the likelihood of relief by 16 percentage points. As the human rights conditions in a country become more repressive, as measured by the U.S. Department of State, relief becomes more likely. Shifting those conditions from the best possible score to the worst increases the likelihood of relief by 18 percentage points. The combined effects of the applicant characteristics in this truncated sample are heartening, as they serve as a robustness check on the models we presented in Chapter 3. Here, we see that these characteristics matter in much the same way at the end of our period, 2005 to 2010, as they do over the entire time span of our analysis. Our analysis of how local conditions affect IJ decision making begins with the notion that as economic conditions worsen in a metropolitan area IJs may become less likely to grant relief. We find this to be true, at least for the ideologically average IJ. The entry for the effect of MSA unemployment

Local Conditions

101

captures moving from low unemployment (3.7 percent) to high unemployment (9.4 percent) holding IJ ideology constant at average (roughly 1 on the factor score for asylum liberalism). There is a small decrease, of 2 percentage points, for the average IJ as unemployment increases. As a second measure of the impact of economic conditions, we included a measure of the percentage of overall employment in a metropolitan area in immigrant-laborintensive industries. As this number increases—as these industries become more important to the local economy—there is an 8 percentage point increase in the proclivity of IJs to grant relief to applicants, at least for the average IJ. Therefore, there is some support for the notion that economic conditions matter unconditionally, as we expected a priori. Table 4.3 parses the conditional effects of the economic conditions. The first two rows measure how changing unemployment changes IJ decision making. The first column contains estimates for a conservative IJ while the second column contains estimates for a liberal IJ. For a conservative IJ increasing unemployment increases the likelihood of relief by 1 percentage point, while for a liberal IJ there is actually an 11 percentage point decrease in the probability of relief. Therefore, liberal IJs behave as we would expect to increasing unemployment: they act as though they tend to see applicants as a potential labor threat to other segments of the local population, perhaps particularly those low-skilled segments to whom liberals may be more solicitous. However, conservative IJs react in a manner that suggests that their major concerns about bogus asylum seekers are somewhat mitigated when employment conditions are bad. In other words, conservative IJs seem to view applicants who seek asylum when labor conditions are poor as more credible. Therefore, the effect of local economic conditions seems to be somewhat variable depending on the ideology of the evaluating IJ. The conditional effect of the concentration of industries using high levels of immigrant labor is unidirectional and expected. As this concentration increases Table 4.3. Economic Threat Economic Threat

Conservative

Liberal

Low unemployment High unemployment

0.36 [0.31, 0.42] 0.37 [0.28, 0.47]

0.49 [0.42, 0.55] 0.38 [0.28, 0.48]

Low immigrant employment High immigrant employment

0.33 [0.27, 0.38] 0.40 [0.35, 0.45]

0.37 [0.23, 0.51] 0.53 [0.43, 0.63]

102

Chapter 4

both conservative and liberal IJs become more likely to grant asylum, with liberals reacting particularly strongly to this increase. This finding buttresses our suspicion with respect to the more general effects for unemployment, as it appears as though liberal IJs become significantly less worried about the effects of increasing unemployment on the local workforce when it is clear that there are likely to be jobs for those admitted, probably in industries that are not as attractive to native workers. Contact and cultural threat hypotheses are evaluated with a second set of variables: the percentage of the population that is foreign-born and whether the applicant is seeking asylum in a community with a large matching diaspora community. Evaluating the effect of foreign-born population levels for the average IJ indicates a large negative effect on the average IJ: when the percentage of a population that is foreign-born changes from 7 percent to 30 percent, the ideologically average IJ is 11 percentage points less likely to grant asylum, a result that provides some support to cultural threat theories. However, the presence of a large diaspora community in the area has the opposite effect, increasing the likelihood of asylum for an applicant by 4 percentage points when an ideologically average IJ is making a determination. Our investigation of the conditional impact of these conditions is contained in Table 4.4. Here we see that the effect of increasing the foreign-born population in an area is similar across IJ ideologies, although liberal IJs seem to be somewhat more responsive to these changes than are conservative IJs. The effect is negative, meaning that as the foreign-born population in an MSA increases, both conservatives and liberals become less likely to grant asylum. This implies that some degree of cultural threat may matter for IJ decision making, although whether that effect is directly upon IJs or is instead a strategic reaction we cannot discern. The conditional effects for diaspora communities are positive and indicate that conservative IJs in particular are more likely to grant asylum in the presence of a large diaspora community—although liberal IJs also experience a small uptick Table 4.4. Contact and Cultural Threat Contact and Threat

Conservative

Liberal

Low foreign-born population High foreign-born population

0.40 [0.36, 0.44] 0.35 [0.31, 0.39]

0.53 [0.42, 0.64] 0.43 [0.33, 0.53]

No diaspora community Diaspora community

0.35 [0.32, 0.39] 0.40 [0.35, 0.44]

0.46 [0.39, 0.53] 0.46 [0.37, 0.56]

Local Conditions

103

in their propensity to grant relief. What we observe then is rather complex. Though the level of foreign-born in an MSA seems to depress the likelihood of a grant, the presence of high concentrations of co-ethnics in a diaspora community seems to somewhat countermand this threat effect, suggesting that a generic threat might be counteracted when it is plausible that the applicant is more likely to find social support networks in the area. Finally, we set out to investigate Hopkins’s (2010) provocative theory of politicized places. Ignoring the constituent terms of the salience of immigration nationally and the change in the foreign-born population between 2000 and 2005, we focus our attention on the interaction term for those two measures. Here we see that the effects on the average IJs are the same as those Hopkins finds for local populations. The effect is substantively slight because, for the average IJ, the rate at which the foreign-born population is changing in the area has a negative effect on grant likelihood, while increasing national immigration salience tends to increase the likelihood of a grant. Reactions to politicized places may be conditional on IJ ideology—a possibility we investigate in Table 4.5. In the fi rst row we hold the rate of change in the foreign-born population constant at a low level (7 percent change, roughly Charlotte in our time span) and alter the salience of immigration as an issue from low to high. In the last two rows we hold the rate of the change in the foreign-born population constant at a high level (18 percent, roughly Las Vegas in our time span) and again alter the salience of immigration from low to high. The reactions of IJs to politicized places are highly conditional on their ideology. Conservative IJs become less likely to grant asylum when a place becomes politicized, whereas liberal IJs become more likely to grant asylum when a place becomes politicized. The evidence suggests, then, that conservative IJs react to the politicization of immigration in a community in ways that are consistent with the masses studied by Hopkins. However, liberal IJs act in a way that is inconsistent with Hopkins’s evidence for the public, a reaction about which we are willing to speculate. Examining the constituent terms of the three-way interaction upon which Table 4.5 is based provides some clues: it appears that IJs react more strongly to the framing of immigration as an issue in the national media than they do to the rate of change of the foreign-born population in an area. Therefore, for some reason the divergence in the reactions of IJs to the politicization of immigration in their communities might be premised on their differential reaction to the framing of the issue nationally, and less so on the actual rate of change in the foreign-born population.

104

Chapter 4

Table 4.5. Politicized Places Politicized Places

Conservative

Liberal

Low salience, slow change High salience, fast change

0.36 [0.33, 0.40] 0.33 [0.26, 0.40]

0.43 [0.35, 0.52] 0.60 [0.35, 0.85]

More evidence on how local political elites who do not face electoral reprisal react to politicized places is needed.

Conclusion We find evidence that IJs are responsive to local conditions, largely in highly conditional ways. Put differently, liberal and conservative IJs react differently to the same set of local conditions, so it is not just that as the conditions in a locality change so too will the chances of an applicant gaining asylum, but that within a community whom the applicant draws as judge will also influence how those local conditions affect the chances of relief. The results can be viewed in two different ways. First, for the average IJ, economic conditions matter as we would expect, with high unemployment decreasing the likelihood of relief, while a large industrial base employing immigrants boosts the chances of relief. Th reat seems to dominate the decision making of the average IJ, with the presence of a particularly large concentration of co-ethnics mitigating the perception of threat. Second, conditional reactions are strong with respect to both unemployment and the politicized places theory of immigration backlash, with conservative and liberal IJs reacting differently to the same conditions. We also find strong evidence that the number of registered NGOs in an area can have a profound effect on the likelihood that an applicant will be granted asylum. This fact will tend to increase the apparent randomness of the asylum adjudication, as there is a rather large disparity in the distribution of these NGOs across the country. Unsurprisingly, most are concentrated in large coastal cities where large numbers of asylum applicants appear. Indeed, in our data, 35 percent of the registered NGOs are located in just four cities: Chicago, Los Angeles, New York, and Washington, D.C. Asylum seekers in these cities have a substantially better chance of avoiding return than do those who apply in cities with fewer groups, such as Cleveland

Local Conditions

105

or Memphis, where there are fewer than ten registered NGOs and the number of applicants is comparatively large. This evidence, particularly when coupled with our data on diaspora communities, suggests that IJs themselves are not the only source of major disparity in the asylum process—the amount of support available to potential asylum seekers is also, apparently, highly variable and matters in the decision-making process. Both diaspora communities and the presence of asylum-related NGOs may provide opportunities for support and better representation in the courts. For example, a large diaspora community may be a source of fi nancial support to the asylum seeker (who is unable to legally work). Th is may also ameliorate concerns about the cost of the new arrival on the community at large. In terms of credibility, if an asylum seeker has a support system, presumably there are more individuals who can indirectly testify regarding the asylum seeker’s fear of return. Both may also indicate a higher likelihood of the asylum seeker having potential country experts available. There may also be a lower likelihood of culturally specific behaviors being misinterpreted by the IJs, which may undermine credibility findings. In any case, the responsiveness of IJs to local conditions is both surprising and unsurprising. It is surprising from the perspective of the law, which does not tend to view these local conditions as relevant for asylum decisions, yet the responsiveness of IJs clearly affects decisions on the granting of relief. This is less than surprising given the political realities faced by IJs, who have more limited tenure protections than federal judges and who have also, in limited circumstances, been found responsive to local conditions. In their responsiveness to local conditions, IJs perhaps resemble street-level bureaucrats more than they do judges, and thus theoretical approaches to their decision making that focus on them from a bureaucratic perspective instead of a judicial perspective are likely to yield additional insights into their decision making. More data and analysis will also be able to shed light on whether the effects for local conditions that we have uncovered here are direct, in the sense that they tend to alter IJ attitudes toward asylum applicants, or whether they are instead strategic, in the sense that IJs anticipate the negative (or possibly positive) reactions of the community and trim their sails accordingly. In the next chapter we shift our focus by examining asylum seekers’ decision on whether to appeal an unfavorable IJ decision and examining factors that influence the Board of Immigration Appeals’s decisions.

CHAPTER 5

Appealing to the Board of Immigration Appeals

In this chapter we focus on the relationship between asylum applicants, immigration judges (IJs) and the Board of Immigration Appeals (BIA). The Board, which is a part of the Executive Office for Immigration Review (EOIR), is the nation’s chief administrative body for immigration law and thus serves as the highest administrative tribunal adjudicating immigration matters. It has responsibility for interpreting and applying immigration laws nationally. For many asylum seekers, the BIA has historically been their last, best chance to challenge a final deportation order. For IJs the BIA is the most immediate venue of review in which their decisions face a real possibility of being overturned or remanded. As we will see, this historical pattern has changed somewhat with reforms in the Department of Justice. Now IJs are, because of these changes, considerably less likely to have their decisions disturbed than they were previously. Our interest here is in how the behavior of the BIA might alter the decision making of the IJs and asylum applicants. But subsidiary to that interest, we need to understand why cases are appealed to the BIA and how outcomes are determined on appeal. The BIA was constituted by a directive of the attorney general in 1940 and “first began to resemble its current form” under regulations implementing the INA of 1952, “which gave the Board appellate jurisdiction to review final deportation and exclusion decisions of immigration judges (then known as Special Inquiry Officers), as well as certain other decisions of immigration enforcement officials” (Arnold and Porter LLP 2010, 3–4). The board was placed within the DOJ, in the newly created EOIR, in 1983. Regulations give the authority to appoint Board member who are to “act as the Attorney General’s delegates in the cases that come before them” (8 C.F.R.

Board of Immigration Appeals

107

§1003.1 (a)(1)). In addition to appointing its members, the AG determines the structures and procedures of the BIA. The BIA has broad authority to review decisions of IJs and does so through a paper review of “the record below, which includes transcripts of testimony, exhibits, briefs submitted by the parties, and the written decision of the immigration judge” (Rana 2009, 843). Interestingly, Department of Justice (DOJ) regulations give both applicants and the Department of Homeland Security (DHS) the right to appeal IJ decisions to the BIA (Legomsky 2010, 1643). Under current regulations, IJ decisions are subject to the reasonableness standard of appellate review, which requires that the BIA find that the IJ reached an unreasonable conclusion in order to reverse a decision. The facts of a case are reviewed for clear error by the BIA; whereas, application of the law is reviewed de novo. Board decisions may be referred to the attorney general for review either at the attorney general’s request or at the request of DHS. The AG may vacate any BIA decision and instead issue his or her own decisions, although this power is rarely used (Arnold and Porter LLP 2010, 3–5). Applicants for relief may seek judicial review of final agency decisions in deportation hearings by going directly to the appropriate federal circuit court. The government does not appeal decisions of the BIA to the circuit courts, as the BIA is presumed to speak for the government vis-à-vis the attorney general. Legomsky (2010, 1649n64) explains this as being “because the United States presumably cannot bring an action against itself” and notes that regardless “in practice, the government has no need to ask a court to reverse a BIA decision because the attorney general can simply do so unilaterally.” Current regulations allow for fifteen Board members, but the size of the board has varied considerably overtime. It began with five members but increased under the Reno DOJ to nine in 1994, twelve in 1995, fifteen in 1996, eighteen in 1998, and twenty-one in 2000, with the final regulations increasing the size to twenty-three going into force in 2001 (see Arnold and Porter LLP 2010, 3–5n28). Some scholars, such as Schoenholtz (2004–5) and his coauthors (RamjiNogales, Schoenholtz, and Schrag 2009), claim that the BIA has been the “single most important decision maker in the immigration system” (Schoenholtz 2004–5, 353). Furthermore, Schoenholtz argues that because the BIA reviews cases on a national basis and sets precedents under which IJs and asylum officers (AOs) are bound, and because the Supreme Court rarely issues asylum-related decisions, “the BIA essentially interprets immigration law for the nation” (353). The role of the BIA in the asylum process has changed

108

Chapter 5

significantly over the past decade and a half, as three attorneys general have engaged in a series of structural and procedural reforms, some of which have politicized the BIA and some which of seem to have diminished its decisional independence. The reforms have diminished the BIA’s role in immigration adjudication and have arguably shifted the BIA’s workload (and concomitant costs) onto the federal circuit courts. The reforms have also hurt perceptions of the BIA’s legitimacy and have led to calls for reforms that would abolish it. In the sections that follow, we discuss the series of reforms that have taken place under Janet Reno (1993–2001), John Ashcroft (2001–5), and Alberto Gonzales (2005–7), focusing especially on the far-reaching reforms of Ashcroft that changed a half century’s process and norms and, more important, arguably politicized both the structure and outputs of the BIA. We also examine the implications of these reforms for asylum seekers and the broader immigration adjudication system. We then turn to a theoretical exploration of the applicant’s decision to appeal an IJ decision and test the hypotheses we derive. Next, we turn to the BIA’s decision making, explicating and subsequently testing our theoretical expectations. Finally we explore the importance of our newfound understandings for IJ decision making.

Reforms of the BIA Across Three Attorneys General The push for reform of the BIA came in response to the “ ‘crushing backlog’ of immigration appeals, which had increased nearly nine-fold since 1984,” going from three thousand cases received in FY1984 to almost thirty thousand in FY2000 (Rana 2009, 843 and citations within). The increase over time appears to have come from an increase in the number of appealable IJ decisions, which is likely the result of “record migration and a huge increase in expulsion (from 30,039 expulsion proceedings in 1990 to 185,731 in 2000) and an increase in the rate at which those decisions were appealed” (Palmer 2006–7, 17n18 and citations within).1 In addition, the increase is also associated with two phenomena occurring at the end of the 1980s and in the early 1990s: the “increased enforcement of sanctions against employers who hire undocumented aliens” and the “fact that employment authorization could be quickly obtained by filing an asylum application” (Palmer 2006–7, 17; also see Martin 1990; Schrag 2000). Palmer notes that “another cause of the BIA’s backlog may have been Congress’s frequent amendments to the INA,

Board of Immigration Appeals

109

particularly the complex changes brought about by the 1996 Antiterrorism and Effective Death Penalty Act (AEDPA) and Illegal Immigrant Reform and Immigrant Responsibility Act (IIRIRA),” and he also points out disagreement among Board members over the cause of the backlog including claims of management problems within the BIA such that during the early 1990s only three of the five permanent seats on the BIA were fi lled (with IJs intermittently filling in on the Board) and that Board members with no background in immigration needed time “to become proficient in their work” (Palmer 2006–7, 17). Three attorneys general enacted a series of reforms to address these issues. For easy reference, the reforms are summarized in Table 5.1. Attorney General Janet Reno responded first. Historically the Board’s procedure had been to decide and write cases in panels of three members until October 1999, when Reno instituted what was later referred to as a “pilot project” and constituted the first round of “streamlining” reforms of the BIA’s appellate review (Rana 2009, 843–44). Under her regulation, the Board chairman was authorized to designate categories of cases as suitable for review by single board members and authorized single board members to affirm IJ decisions in certain circumstances without writing any opinion (Palmer 2006–7, 18), a process abbreviated as AWO (affirmance without opinion). According to the BIA chair’s memoranda to Board members, this streamlining was “intended to apply only to the most ‘routine’ cases where there was no “reasonable possibility of reversible error in the result reached below” and “asylum and other complex cases were specifically exempted from these streamlining rules” (844).2 Schoenholtz (2004–5) refers to these changes as the most significant development of Reno’s governance of the BIA. In addition to these changes in procedures, Reno also increased the size of the Board to twenty-three members to address the Board’s increasing caseload. Ramji-Nogales, Schoenholtz, and Schrag (2009, 61) report that “she added members who had served as INS trial attorneys or Office of Immigration Litigation attorneys at the Department of Justice, a senior congressional staffer who had served as the Republican chair of the House Immigration Subcommittee, and several lawyers from private practice, advocacy and academia.” They then assert that “the latter appointments balanced somewhat the predominant government experience of existing members and of her appointees who had prior law enforcement experience.” As we will see below, Reno’s reforms were somewhat limited compared to Ashcroft’s subsequent reforms; however, they appear to have been successful in reducing

110

Chapter 5

Table 5.1. Major Attorney General Reforms of the BIA (Reno Through Ashcroft) Reno October 1999 Streamlining reforms

Size of Board membership

Board chairman authorized to designate categories of cases as suitable for review by single board members Authorized single board members to affirm IJ decisions in certain circumstances without writing any opinion Increased size of the Board to 23 members Ashcroft August 2002

Streamlining reforms

Size of Board membership

Made single-member decisions the dominant means of adjudication for the overwhelming majority of cases Made single-member summary affi rmances commonplace Effectively eliminated the BIA’s de novo review of factual issues Required the Board “to defer to the Immigration Judge unless a decision is clearly erroneous” Note: In May 2002 the BIA chairman continued the expansion of single-member AWOs in essentially all cases Reduced size of Board from 23 members to only 11 Gonzales August 2006

Performance and quality

Announced 22 measures/reforms he deemed necessary to improve the performance and the quality of adjudication Annual performance evaluations

much of the BIA’s backlog. Rana (2009, 844) reports that “by 2001 the Board was already deciding and disposing of more appeals than it was receiving and had greatly cut its backlog.” An independent audit by Anderson LLP, released in December 2001, reported “a fi ft y-three percent increase in the overall number of cases completed,” found that there was no indication of “an adverse effect on non-citizens,” and “concluded that the new streamlining rules resulted in an unqualified success” (Schoenholtz 2004–5, 354).

Board of Immigration Appeals

111

As Schoenholtz (2004–5, 354–55) points out, despite the success attributed to Reno’s reforms in the December 2001 report, Ashcroft announced a series of new policies in the name of streamlining a mere two months later, which “fundamentally changed the nature of the BIA’s review function” and “radically changed the composition of the Board.” Rana (2009, 844–45) further notes that “though the Board’s backlog was already shrinking, and the Board had already increased its productivity by 65 percent since 1998,” Ashcroft’s expanded streamlining provisions “effectively restrict[ed] review for nearly all appeals to the Board.” According to Schoenholtz the new policies resulted in three major changes, which we will deal with in order below: (1) making single-member decisions the dominant means of adjudication for the overwhelming majority of cases and making single-member summary affirmances commonplace, (2) effectively eliminating the BIA’s de novo review of factual issues, which “established the primacy of immigration judges as factfinders,” and requiring the Board “to defer to the Immigration Judge unless a decision is clearly erroneous,” and (3) reducing the membership from an authorized twenty-three members to only eleven (253–55; see also Rana 2009; Palmer 2006–7; Legomsky 2010; Alexander 2006). These changes likely increased the importance of IJs in the process, as their decisions in asylum cases are more likely than ever to be final. Not only did the attorney generals cite the backlog as a reason for the changes, he also “sought to justify the drastic limitations on review by citing heightened national security concerns stemming from September 11, explaining that this ‘reorganization’ was part of his plan for ‘protecting America from terrorist attack’ ” (Rana 2009, 844). Rana (2009, 845) concludes that “with these somewhat fl imsy underpinnings, the streamlining reforms became a part of the Attorney General’s war on terror—and the Attorney General saw unrestrained authority over immigration as a critical weapon in this war.”3 Palmer (2006) notes while the attorney general’s new regulation was still in their notice and comment period, “the BIA Chairman added two large categories to the list of appeals eligible for single-member AWOs: (1) cases involving claims for asylum, withholding, and CAT relief; and (2) cases involving claims for suspension of deportation or cancellation of removal.” The BIA chairman continued the expansion in May 2002 by “designat[ing] as eligible for single-member AWOs all cases involving appeals of IJ or Immigration and Naturalization Ser vice decisions over which the BIA had jurisdiction—essentially all cases” (19). The AG’s new regulation was finally published in August 2002, codifying and further expanding the

112

Chapter 5

streamlining procedure. Rana (2009, 832–33) highlights the breadth of these reforms and their implications for alien applicants: The Attorney General went so far as to authorize individual Board members to dispose of cases through the affirmance without opinion procedure even if there were errors in the immigration judge’s decision below, and even if the Board member did not agree with the reasoning of the decision below. The streamlining rules expressly specified that an affirmance without opinion by a Board member only affirmed the results, not necessarily the reasoning, of the immigration judge’s decision. Rana then contrasts the process of overturning an IJ’s decision and granting relief from deportation, “Board members were required to write a reasoned opinion. Yet a Board member who wished to overturn an immigration judge’s decision would have to expend increasingly limited time and resources to do so” (833). In addition, Ashcroft’s reforms eliminated the Board’s traditional “authority to conduct de novo fact finding and thus overrule the factual findings, and in certain cases the credibility determinations of immigration judges, without requiring remand” (Arnold and Porter LLP 2010, 3–4). The new regulations removed this authority, lowering the standard of review to a “clearly erroneous” standard of review. At the time Ashcroft announced the new regulation, he also imposed strict deadlines which had the effect of each Board member having to review and decide nearly 4,000 appeals a year (Rana 2009, 833).4 Furthermore, he “required BIA members to clear their current backlog of cases of approximately 55,000 cases within 180 days.” This meant that “each member had to decide 32 cases every work day or one every fifteen minutes” (Schoenholtz 2004–5, 357). Rana goes so far as to argue that “under such a system, the agency could decide to deny an immigrant’s claim, rather than grant relief, on the grounds that the Board simply lacked the time or inclination to spend its resources writing a reasoned, public opinion for that par ticu lar case” (836). Before we further examine the criticisms of the procedural reforms and their actual effect, we turn fi rst to the changes in the structure of BIA, as we believe the effects of procedural and structural changes are intertwined. Ashcroft’s decrease of the Board’s membership has received significant criticism given the nature of the Board’s backlog and its relative success

Board of Immigration Appeals

113

under the Reno reforms in reducing the backlog. But more importantly his subsequent action of targeting liberal members for removal has brought claims he was deliberately politicizing the BIA. Schoenholtz (2004–5, 356) reports that “according to the Federal Register, the Attorney General based this determination ‘on judgments made about the historic capacity of appellate courts and administrative appellate bodies to adjudicate the law, the ability of individuals to reach consensus on legal issues, and the requirements of the existing and projected caseload.’ ” Palmer (2006–7, 24) reports that the DOJ (in its supplemental information accompanying the 2002 regulation) claimed that Reno’s expansion of the BIA “had degraded, among other things, the cohesiveness and collegiality of [its] decision-making process, and . . . the uniformity of its decisions” and that the BIA’s precedent decisions “indicated an inability to reach consensus on even fundamental approaches to the law.’ ” Ashcroft then justified the reduction in the number of Board members, claiming it should “increase the coherence of Board decisions, and facilitate the en banc process, thereby improving the value of Board precedents” (Palmer 2006–7, 24; see also Alexander 2006, 12). Critics such as Schoenholtz (2004–5, 356–57) argue that in the actual downsizing Ashcroft targeted the “newer members [appointed by Reno] who came from the practice of asylum and immigration law, advocacy, and academia.” Ramji-Nogales, Schoenholtz, and Schrag (2009, 62) concur, noting that the members Ashcroft removed were five members appointed by Reno, not those with the least seniority but instead those members on the Board who “most disagreed with him ideologically.” Similarly, Baum (2010, 1524) concludes, “[B]ased on these members’ decisional tendencies, it appears that they were reassigned largely due to their tendency to favor appellants more than their colleagues did” (see also Levinson 2004, 1157–58). Schoenholtz (2004–5, 357) reports that the executive director of the Center for Immigration Studies (a nonprofit organization that advocates immigration reduction) “endorsed the removal of these members—observing that ‘Board members should clearly represent the attorney general’s views, since they are carry ing out his responsibility.’ ” Our assessment of the five members’ ideology (based on their career socialization scores; see below) suggests that Ashcroft did in fact target some of the most liberal Board members on the bench. As we will discuss below, the scores range from 0 to 3.3, with an overall mean of 0.67 in our twentyyear time frame. The scores of the Board members removed by Ashcroft are as follows, in ascending order of liberalism: Paul Schmidt (1.113, which is

114

Chapter 5

the cutoff for the 75th percentile), Gustavo Villageliu (1.596), John Guendelsberger (2.035), and Noel Brennan and Claudia Espinoza (both at 3.334). These five members are five of the nine most liberal members to have served on the board in our time frame, including two of the most liberal. By comparison, Board member Garry Malphrus, who was one of the politicized Goodling-approved BIA candidates who eventually made it to the board under new hiring policies, scores 0.199 on our scale.5 These scores, along with Levinson’s analysis, rather strongly support the assertion that the Ashcroft administration was attempting to alter the Board’s ideological composition. Of course, these five appointees are also all Reno appointments, and as we will see below there is a huge shift in the liberalism of appointees to the Board under Reno. Many scholars worry about the impact of the seeming attempt to ideologically cull the Board on the members’ and IJs’ decisional independence (for example, Levinson 2004; Palmer 2006–7; Legomsky 2010). Legomsky (2010, 1668) specifically worries about “the erosion of the immigration judges’ and BIA members’ job security and the real and perceived effects of that erosion on their decisional independence,” noting that when Ashcroft announced his plan he gave no details as to the criteria he would apply in removing Board members. Former congressional counsel Peter Levinson (2004) argues that “the Attorney General succeeded in moving the [BIA] in a conservative direction just by announcing the downsizing plans—and the result of the downsizing was to remake the Board into a largely homogeneous body without dissent.” 6 Legomsky’s (2010, 1670) final assessment is more disturbing: “[T]he postpurge outcomes in BIA cases have been significantly less favorable to immigrants.” Overall, academic reviews of the reforms effects are quite consistent, and they strongly suggest that the net result of the reforms has been detrimental to aliens appealing IJ decisions while also tremendously increasing federal circuit court caseload. The most obvious effect of the Ashcroft changes has been a huge increase in the number and proportion of BIA decisions that are now appealed to the federal circuits. In 2001 appeals of all BIA decisions constituted only 3 percent of all federal appeals filed—in 2002 the percentage increased to 10.9 percent and then it increased further to 14.4 percent in 2003 (Hethmon 2006, 1002).7 Ramji-Nogales, Schoenholtz, and Schrag (2009, 65) report that in February 2002 approximately two hundred BIA cases per month were to appealed the circuits, whereas with one year the

Board of Immigration Appeals

115

number had grown to nine hundred per month and by April 2004 the number had grown to more than one thousand. Ashcroft’s reforms clearly have overwhelmed the federal circuit courts, which now have to duplicate much of the BIA appellate review, and created a crisis in some circuits such that one circuit had to eliminate oral argument in asylum appeals (Legomsky 2010, 1646–47).8 In addition, the reforms have had two significant substantive effects on the BIA—changes we assume were intended by the AG. First, the reforms have changed the type of decisions handed down by the Board. For example, Ramji-Nogales, Schoenholtz, and Schrag (2009, 66) find that the percentage of cases that are multimember panel decisions dropped from over 70 percent in 2002 to less than 10 percent in 2003. In addition, Legomsky finds that the percentage of cases that were AWOs rises from 6 percent in FY2001 to 31 percent in FY2002, 36 percent in FY2003, and 32 percent in FY2004. These changes are substantively significant. As Legomsky (2010, 1663– 64) argues, a reasoned opinion is valuable because “it requires the adjudicator to consider the arguments of the losing side with care. . . . [T]he very process of writing an opinion forces adjudicators to confirm that their tentative conclusions are the ones most compatible with the evidence and the law.” The dominance of AWOs very likely affects IJ decision making as well. Legomsky (2010, 1664) argues that without the appropriate guidance to bind IJs and DHS officials, they will have to “speculate about the BIA’s likely future interpretations” and “speculation can only increase the number of reversals and, and spawn inconsistent treatment of similar cases,” and furthermore, “without reasoned opinions, it becomes easy for appellate adjudicators to base their decisions, consciously or unconsciously, on visceral reactions that reflect their own political outlooks, thus further eroding both accuracy and consistency.” In addition, he argues that “because writing a persuasive opinion takes time, BIA members with staggering caseload pressures and far too little time per case have a strong incentive to affirm rather than reverse” and “consequently, and not surprisingly, the percentage of cases in which the BIA reversed immigration judge decisions dropped precipitously after the 2002 reforms” (1662). This fact raises concerns given the well-documented disparity in grant rates among IJs and the possibility of error in IJ decision making. The switch from panel decisions also raises substantive concerns. As a host of judicial behavior scholars note, panel decisions diff use subjective

116

Chapter 5

biases, encourage deliberation, promote consensus, minimize inconsistency, and permit dissenting opinions, which can be important in shaping future law (e.g., Kastellec 2013; Farhang and Wawro 2004), all of which is even more important given that so many immigration cases are argued pro se and without legal briefs (Legomsky 2010). In addition, the reform has raised concerns that since the Board can designate as precedents only decisions that were in en banc and multimember, the number of published precedents has diminished, which “limits the Board’s traditional function of crafting greater uniformity in immigration law” (Arnold and Porter LLP 2010, 3–3, 3–4). Empirical evidence suggests that these concerns are valid—several studies demonstrate that following these changes, the results of BIA decisions have turned against the alien applicant. For example, Dorsey and Whitney LLP (2003, 235) found that between June 2000 and October 2002 the grant rate decreased from 25 percent to 10 percent. Furthermore, Palmer (2006–7, 23) reports that the BIA rejected 86 percent of its appeals in October 2002 as compared 59 percent the previous year.9 Similarly, there was a 70 percent decrease in the success of represented applicants and a 77 percent decrease for unrepresented applicants and an 84 percent decrease for unrepresented applicants from non-asylum-producing countries (68).10 Before we continue, it is important to point out one large caveat to the statistics typically cited in the literature. These numbers are generated from all removals heard by the BIA—even cases not related to asylum. Our data are a much more limited set of all asylum claims decided on the merits. For example, in 2010 the BIA decided 285,892 removal cases, but in that year we have only 19,127 asylum cases reviewed by the BIA. The BIA also hears cases regarding the DHS (or Immigration and Naturalization Ser vice prior to the reorganization) on topics such as visa petitions, waivers, and fines (EOIR 2011, T2). IJs also hear cases other than asylum, including removal, detention, rescission, and so on (EOIR 2011, C2). It is difficult to sort out whether the perceived effects are related to changed ideological composition of the Board or the procedural regulations. Regardless, as Baum (2010, 1527) points out, “the attorney general’s effort to change the pattern of decisions by BIA members was successful.” TRAC concludes that Ashcroft’s reforms resulted in a wave of criticism from the federal circuit courts against IJs and the BIA, and that Attorney General Alberto Gonzales responded in part to this criticism (as well as the emerging

Board of Immigration Appeals

117

Monica Goodling scandal) by ordering the DOJ to undertake a special study of the EOIR (TRAC 2009). On August 6, 2006, Gonzales announced twentytwo measures/reforms he deemed necessary to “improve the performance and the quality” of both the immigration courts and the Board of Immigration Appeals (TRAC 2009). Most important among the reforms was the implementation of annual performance evaluations of IJs and Board members. Gonzales’s memorandum justified the periodic review of “the work and performance” of each IJ and Board member as being consistent with the performance appraisal records elsewhere in the DOJ to “assess the work of personnel at all levels,” and he argued that the “EOIR performance evaluations will allow for identification of areas where an immigration judge or Board member may need improvement while fully respecting his or her role as an adjudicator” (Gonzales 2006). In EOIR’s 2009 report to Congress the EOIR noted that the Board members “were placed on their work plans on July 1, 2008” (EOIR 2009b). TRAC reported that the first full annual review for BIA members was to occur in July 2009 (TRAC 2009). In addition, new Board members, like new IJs, would be assessed during the two-year trial period to “assess whether the new appointee possesses the appropriate judicial temperament and skills for the job and to take steps to improve that performance if needed,” and in addition the EOIR was required to file a report with the deputy attorney general on “the temperament and skills” of each new appointee prior to the end of the two-year trial period. TRAC obtained copies of the generic performance evaluation document for Board members through a Freedom of Information Act request. The document includes a midyear evaluation as well as annual evaluation and identifies three “job elements” that assessed quality of adjudications, accountability for organ i zational results, and productivity. All three elements are identified as “critical.”11 Legomsky (2010, 1662– 63, 1674) argues that the performance evaluations give an incentive to affirm rather than reverse IJs “by emphasizing productivity,” and because immigrants file the overwhelming number of appeals with the BIA (85 percent to 93 percent), the incentive to affi rm means outcomes that favor the government over the alien. The per formance evaluations raise serious questions about the decisional independence of both IJs and Board members (see Legomsky 2010). We want to return more fully to another major event in the DOJ, the socalled Monica Goodling scandal, which may have some implications for the

118

Chapter 5

BIA. The Office of the Inspector General’s investigation revealed that “from 2004 to 2006, high officials from the White House and the Department of Justice had bypassed the usual application procedures to appoint immigration judges based on their Republican Party affi liations or their conservative political views. In all but four cases, the hiring was accomplished without public competition, and more than half the appointees had no prior immigration experience” (Legomsky 2010, 1666). The report highlights a politicized process that included the White House “solicit[ing] candidates for IJ positions from the Republican National Lawyers Association, Republican National Committeemen, state and local Republican Party officials, the Federalist Society, and prominent Republicans” and then “provid[ing] those candidates to Goodling for consideration (U.S. Department of Justice, Office of Professional Responsibility and Office of the Inspector General 2008, 102– 3). The candidates were then vetted through a political screening process by Goodling and her associates (103). This process was apparently applied to potential BIA members as well. The Office of the Inspector General reports that “Goodling also acknowledged in her congressional testimony that she ‘took political considerations into account’ in connection with recommending four persons to be appointed to career positions on the Board of Immigration Appeals and despite advice from the Office of Legal Counsel that the appointments were Schedule A career appointments, she subsequently selected candidates for the four vacant BIA positions based on political or ideological considerations” (110).12 The four candidates were not appointed to the BIA; however, one, Malphrus, was eventually appointed through the new hiring process announced by Gonzales on April 2, 2007.13 Examining the series of procedural and structural changes leaves us with a set of interrelated and important questions that merit empirical assessment. To what extent has the politicization of the hiring of the board’s membership affected outcomes? Has it shifted the board’s decisions against asylum seekers in par ticu lar? To what extent have Ashcroft’s streamlining reforms and Gonzales’s per formance evaluations affected the Board’s outcomes? Have these reforms enhanced or diminished the role of the Board members’ ideology? We seek to answer some of these questions in the empirical analyses of this chapter. First, we turn to the law and courts literature to build a theoretical foundation to establish a more informed basis for assessment. In doing so, we look fi rst at the

Board of Immigration Appeals

119

asylum seeker’s decision to appeal to the board because of a potential selection effect, which we explain below.

The Decision to Appeal Theoretical Expectations

One difficulty that arises in attempting to understand the Board’s outcomes is that there is a potentially important selection effect in appellate decisions that must be considered. If litigants in asylum cases are rational, as we will argue below, then they will tend to appeal decisions that are not representative of the universe of all asylum decisions made by IJs. Thus the body of appealed cases is unlikely to be representative of the population of cases as a whole. For instance, we would expect that almost all of the appealed cases will involve adverse credibility determinations, whereas only a third to half of the cases decided by IJs likely involve such a determination. To deal with this potential bias, we must account for the initial decision to appeal in our statistical models. And to do so appropriately, we first need to explore the theoretical expectations concerning these decisions. In addition, in our period, a rather significant percentage (47 percent) of all IJ asylum decisions are appealed (compared to 8 to 10 percent of all IJ decisions from 2006 to 2010; EOIR 2011, X1), which makes it substantively important to analyze the factors that motivate an appeal. Our theory of the decision to appeal is straightforward: asylum applicants will appeal an asylum decision when it is rational to do so.14 Given the punishment faced by an asylum seeker who has been denied relief of any type—deportation from the country—we believe it is most sensible to analyze the decision to appeal as though it is analogous to the decision of a criminal defendant to appeal. The social science literature on the decision to appeal has focused predominantly on rationality.15 What is meant by rationality in these studies is that the benefit of an appeal exceeds its costs (Landes 1971; Priest and Klein 1984; Songer, Cameron, and Segal 1995; Posner 1985, 1986). This is often deemed the rational actor approach to explaining the decision to appeal. In the simplest terms, the rational actor approach implies that an applicant (or more rarely the government) will appeal a decision when the potential benefits of an appeal outweigh the costs. From a rational actor approach there is

120

Chapter 5

very little reason that an applicant who has been denied relief will not appeal his or her case. Two potential benefits arise from pursuing an appeal, and there is little downside. First, the appeal, even if not successful, allows the applicant to delay ultimate deportation for some time (see Legomsky 2010). Our discussion with immigration attorneys suggests that the delays vary greatly and can be anywhere from five months to a couple of years. One NGO reported that currently appeals to the BIA are taking about two years for nondetained applicants.16 Concomitantly, we make one caveat to this expectation. Applicants who are in detention are less likely to perceive an appeal as providing this benefit as it will prolong their detention (see Palmer, Yale-Loehr, and Cronin 2005, 73–74). Legomsky (2010, 1703) notes that 48 percent of immigrants in removal proceedings are detained, and in our dataset we find that 9 percent of asylum applicants are detained at the time of their hearing, with an additional 14 percent having been detained at some point in the process. Furthermore, he reports that for detained immigrants “the average elapsed time” for BIA appeals is ninety-five days, noting that “former BIA Chair Juan Osuna has noted that BIA policy is to decide all appeals involving detained immigrants within 150 days and has said that, in practice, appeals by detained immigrants have taken an average of ninety-five days to complete” (1701). Second, there is some possibility that an appeal might be successful. One-quarter of all cases that are appealed by applicants result in the grant of some form of relief, so there is a distinct possibility that appealing an IJ’s decision may result in a more favorable outcome. If the applicant faces a strong probability of persecution upon return to his or her country of origin, then the benefit of some degree of relief is even larger. The benefit may be similarly larger for applicants who face return to an authoritarian regime. Thus these applicants may have a stronger incentive to appeal. We believe that rational litigants might consider the ideology of the court to which they are applying, and so we expect that applicants will be more likely to appeal their cases as the liberalism of the BIA increases. The potential cost of an appeal may come in several forms. As we mentioned above, for detained applicants the cost is prolonged detention. The resources required for making an appeal are substantial in terms of time, expertise, and money. While an applicant can appeal pro se, for which the typical fee of $110 is waived, the optional briefing process, which is considered necessary for any probability of winning the appeal, is quite time-

Board of Immigration Appeals

121

consuming. Conversations with asylum practitioners suggest that the appeal process requires three to four weeks for an immigration attorney to complete.17 One attorney reported that he or she was working on a case in which the IJ’s decision was forty-five pages, with a transcript of over four hundred pages. Thus, paying for an attorney represents the biggest cost. We were given estimates of between three and four thousand dollars for a private immigration attorney. Some applicants may be able to secure representation through an NGO. At this point, we have no ability to separate out the different methods of representation. In our BIA dataset, in 83 percent of the cases that are appealed, the applicant is represented. The complicated procedures and the practical necessity of having legal counsel further diminish the possibility that a detained asylum seeker will appeal. Palmer, Yale-Loehr, and Cronin (2005) argue that detained applicants “are likely inhibited from filing petitions for review” because (1) “it is more difficult for them to locate, afford, and meet with counsel” and (2) “many of the detainees who are unable to obtain counsel have a more difficult time preparing their cases pro se” (73–74).18 The initial descriptive statistics support the rationalist perspective. Figure 5.1 displays the rates at which particular types of asylum decisions by IJs are appealed and who filed the appeal. Most important for the basic theory of rationality, applicants appeal when they have lost at the lower level (95 percent of the appeals of denials of any form of relief), and the DHS appeals when the asylum seeker has won the highest level of relief (90 percent of the appeals of IJ grants of asylum). Put differently, when an asylum seeker chooses to appeal, 98 percent of the time that appeal is of an IJ denial of any form of relief, and 65 percent of the time when the government appeals, it is appealing an IJ granting of full asylum. The two intermediate categories of relief are a bit harder to interpret. Asylum seekers and the DHS appear to be about equally likely to appeal a decision of withholding under the Convention Against Torture (CAT), something that might reflect the relative novelty of this form of relief and a lack of precedential decisions on issues likely to arise under the CAT. Alternatively, the middling level of substantive relief granted may simply be an artifact of the few cases in this category. Withholding of removal decisions are predictably much more likely to be appealed by the alien (62 percent of appeals on IJ withholding decisions) than by the government (27 percent of appeals on IJ withholding decisions) given the gulf in substantive benefits available to an alien who receives asylum versus one who receives

122

Chapter 5 Total Cases:

Not Appealed:

506,580

269,568 (53% of total)

Appealed:

Appealed by Both or 3rd Party:

237,012 (47% of total)

10,052 (4% of total appeals)

Appealed by Alien: 216,422 (91% of total appeals)

Appeal of No Relief:

Appeal of Withholding:

Appeal of Asylum:

213,353 (98% of all alien appeals)

2,562 (1% of all alien appeals)

507 (