Refugee Roulette was co-authored by Professor Jaya Ramji-Nogales of Temple University Beasley School of Law and Professors Andrew Schoenholtz and Philip Schrag of Georgetown University Law Center. The authors originally published the study in 2007 in the Stanford Law Review. This book contains the article, with minor updates. It also includes shorter articles from other legal experts about the main study.
The Refugee Roulette authors limited their study to nationals from "Asylee Producing Countries" (APCs), those who had at least 500 claims in FY 2004 and received at least a 30% grant rate. The 15 APCs include Albania, Armenia, Cameroon, China, Colombia, Ethiopia, Guinea, Haiti, India, Liberia, Mauritania, Pakistan, Russia, Togo, and Venezuela. It excludes countries whose nationals received low grant rates, such as El Salvador and Guatemala, as well as those who entered the asylum system for purposes other than to obtain asylum, such as Mexico.
The study defines judge/decisionmaker as deviating from the mean grant rate if his grant rate was 50% higher or lower than other adjudicators in the same office or court. The study employs a regional rather than national standard to account for differences in the composition of immigrant petitioners before each court. The statistical analysis and comparisons assume that clerks at Asylum Offices and Immigration Courts assign cases to adjudicators on a random basis.
Here are some of the most important findings from the main article:
Overall, during FY1999-2005, Asylum Offices had a grant rate of 35%, referring most other cases to Immigration Judges. The referrals included cases in which the petitioner 1) did not appear for his interview; 2) did not meet his burden of proof; 3) did not allege facts sufficient for protection under the statute; or 4) had not filed within 1-year or show "extraordinary circumstances" justifying a delay. Around 7% of cases were dismissed because the petitioner already had lawful status in the U.S. However, the study found considerable variation in grant rates within and between offices. Most officers granted asylum at a rate of 25-50%. Region D produced the most consistent results, with only one out of 64 judges deviating from the office mean by more than 50%. By contrast, in Region H over half of all officers deviated, and five deviated by 130-190%. The study found that disparities persisted even for petitioners of the same nationality. The grant rate for the 290 officers who handled more than 100 cases involving Chinese petitioners varied from 0-90%. In Region H, 31 out of 52 officers who decided more than 25 cases involving Chinese petitioners deviated by more than 50% from the office's mean. Other regions, such as Region C, show relative consistency for most APCs, but wide variation for Indian applicants.
In theory, because Immigration Courts review Asylum Office decisions de novo, they should not match the inconsistency among Asylum Offices. While the overall grant rate for APCs was 40%, the Refugee Roulette authors found serious disparities among courts for six out of 15 APCs from January 2000 to August 2004. In Los Angeles, 32% of judges deviated by more than 50% from the office's mean rate of 41%. Within courts in Los Angeles, Miami, and New York, 8 judges were 50% above the mean and 16 below it - or 32% of all 74 judges deviated substantially from the court's average. Again, variations arose even when holding nationality constant. Chinese petitioners faced the widest variation, with grant rates ranging from 76% in Orlando, 47% nationally, and 7% in Atlanta - in other words, the odds of a Chinese petitioner winning was 986% greater in Orlando than Atlanta.
The authors believe that that several variables contributed to the variation found among Immigration Courts. Approximately one third of applicants come to court without legal representation. Those with representation receive favorable decisions in around 45.6% of cases, whereas for unrepresented plaintiffs the win rate falls to 16.3%. Law school clinics, pro bono firms, and NGOs, which can dedicate more time to case preparation and documentation, win at even higher rates. Likewise, applicants who claim dependents win 48.2% of cases, compared to 42.3% for lone immigrants.
The Immigration Judge's biographical characteristics also correlate with grant rates. Male judges granted asylum in only 37.3% of cases they heard, compared to 53.8% for female judges. Based on prior literature, the Refugee Roulette authors suggest female judges were more likely to have experienced sex discrimination in the past and thus be more sympathetic to immigrants. Likewise, female judges granted asylum to represented petitioners in 55.6% of such cases, compared to a mere 14.3% for male judges.
Refugee Roulette concludes with several observations and policy recommendations. The authors believe the most difficult part of asylum adjudication is determining the credibility of immigrants. Many judges may possess preconceived notions or skepticism based upon their prior work experience or gender. The report cites studies showing that judges with a heavy caseload rely more upon their intuition and bias than reasoned law to make judgments. Dr. Stuart L Lustig used the Copenhagen Burnout Inventory on Immigration Judges and reported stress levels and burnout higher than any other professional group of respondents. As such, Immigration Judges may simply be substituting bias and intuition when they find themselves unable to apply the law or assess credibility in so many cases. The Refugee Roulette authors recommend hiring more judges and law clerks to relieve the burden.
The authors also propose a return to the 1999 streamlining reforms, utilizing 3-judge panels on the BIA and written opinions, rather than Ashcroft's 2002 streamlining. For the Court of Appeals, the authors suggest Congress amend the Immigration and Naturalization Act to allow a "substantial evidence" standard of review over BIA decisions, rather than the current restrictive standard holding that "the administrative findings of fact are conclusive unless any reasonable adjudicator would be compelled to conclude the contrary." However, the authors disagree with the GAO about the need to deploy Assistant Chief Immigration Judges and for supervision of Immigration Judges. After the BIA firings in 2002 and the revelations about the Immigration Judge hiring process in 2006, the authors worry that more administrative supervision might simply lead to political manipulation. Rather, the authors ultimately recommend Congress transfer all immigration adjudication to a new Article I Immigration Court.
Even though the study was overall great, I had a few criticisms:
Even though Refugee Roulette focuses on variations and disparities in asylum adjudication, the study never establishes a baseline for how much variation is normal or tolerable. Given that judges in the U.S. legal system are appointed by politicians from competing political parties, it seems natural that there would be some variation due to ideological or jurisprudential differences. During the 2008-2009 Supreme Court term, almost a third of cases were decided by a 5-4 majority - in other words, half the justices varied from the court's mean. Refugee Roulette begins to address this concern, but does not go far enough. The authors refer to Richard Revesz' study of Courts of Appeal judges voting patterns in environmental cases. Revesz found that during some periods Democratic judges were 50% more likely to vote for an environmental challenge than his Republican colleagues. Likewise, a Republican judge was 100% more likely to vote in favor of industry challenges to EPA regulations. Given Justice Scalia's notorious comments about environmental laws before he joined the Supreme Court, such biases certainly should not shock experienced lawyers.
While Refugee Roulette's methodology is generally sound, its emphasis on a 50% deviation from the mean might hide important variation below that level. Grant rates below the 50% deviation mark yield important information about whether there is a gradation among judges, or whether those deviating are in fact extremes. For example, the authors report that only one officer out of 64 in Region D's Asylum Office deviated from the office mean by over 50%. However, Figure 2.2 shows a relatively wide gradation of grant rates, with several other officers deviating from the mean by 40%, 30%, 20%, or 10%. In fact, for that office, few officers actually had grant rates near the mean. Focusing on deviation from the mean relies too heavily on an arbitrary number (50%), suggesting that a court in which all judges deviate from the mean by 40% suffers from less serious disparities than one in which 10% of judges deviate by 50%. In fact, it seems that grant rates near the mean, rather than deviating by double-digits, are the outlier. Refugee Roulette would have benefitted from more consideration of the spread and gradients of grant rates, not just the outliers.
The Refugee Roulette authors dismiss alternative explanations for the variation they reveal a bit too quickly. The authors acknowledge that different parts of the U.S. might receive immigrants from different countries or even sub-national ethnic minorities, but refuse to explore this possibility further. In a few examples Refugee Roulette does hold nationality constant, particularly for Chinese, to show that discrepancy persists both within and between courts. However, this test would not account for sub-national differences in religion or ethnicity, such as Tibetan or Han Chinese, that could play a more important role in asylum hearings than nationality. Admittedly, as the authors note, this data is difficult to find and code.
Refugee Roulette only spends one chapter discussing possible policy solutions, so predictably its analysis is not exhaustive. It spends several pages attacking "straw men," including "some [who] may suggest..." that reform isn't necessary or that the disparities are acceptable. However, the study never cites "some" people who actually hold those extreme positions. No immigrant advocacy group really believes that EOIR should require judges to fill quotas for granting a certain number of asylum petitions. On the other hand, Refugee Roulette does not explore several serious alternative proposals that some scholars have advocated, such as placing Immigration Judges under the Administrative Procedure Act. Thus, the authors lose a valuable chance to justify their proposals and reject others.
More importantly, Refugee Roulette does not effectively link the authors' proposals to the problems they describe. Many of the proposals, such as the need to grant Immigration Judges greater independence, appear to merely react to excesses under the Bush administration. While the authors believe converting the Immigration Courts into an Article I court would ameliorate this problem, it could create new ones. Instead, moving poorly trained and possibly unprofessional judges into an Article I court could simply insulate them from any supervision or expose them to partisan debates over immigration when Congress votes on reappointing them.