Improving the Exercise of the Attorney General’s Immigration Referral Power: Lessons from the Battle over the “Categorical Approach” to Classifying Crimes

Alberto Gonzales and Patrick Glen have provided a major service in their comprehensive summary and analysis of those decisions in the immigration field rendered personally by Attorneys General over more than seven decades.1 The basic Attorney General review procedure, created by regulation, is known (or actually little known) as “referral.”2 It allows this Cabinet-level official to serve, on selected occasions, as the highest administrative tribunal in formal removal proceedings, usually after a ruling by the Board of Immigration Appeals (“BIA”). Although I will take issue with a few of their assertions, Gonzales and Glen are quite right to highlight the potential value of this procedure, judiciously employed, for setting policy and providing authoritative guidance.3 I begin by expanding on the importance and propriety of the Attorney General’s role in light of the interpretive authority that agencies enjoy under the Supreme Court’s Chevron doctrine.4 Indeed, Cabinet-level head-of-agency review can be particularly appropriate here, in view of the potential linkage between immigration policy and foreign policy. I then draw on a critical history of one controversial Attorney General decision to suggest refinements that could make referral more reliable and enhance its acceptance and legitimacy.

I.     Introduction and Background: The Importance of Chevron

The referral procedure has drawn its share of criticism, especially from immigrant advocates.5 In the view of its strongest critics, referral allows for the infiltration into adjudication of improper influences through the assignment to a political appointee of the final administrative resolution of what should be seen as legal rather than policy issues.6

Such criticism, however, is at odds with the Supreme Court’s view of administrative authority, as reflected most importantly in the Chevron case and later decisions that explicate it.7 When statutes have gaps or ambiguities, Chevron teaches that Congress has implicitly delegated to the administering agency, not the judicial branch, the authority and responsibility to supply a controlling interpretation, within broad bounds of reasonableness.8 Choosing among competing interpretations, each of which would be consistent with the statutory text, quite properly calls for a policy choice. Moreover, the Court has been clear that the agency may reconsider the wisdom of its policy on a continuing basis and, with proper explanation, may alter its interpretation over time.9 The Court has even indicated that an alteration can be appropriate based not only on changed factual circumstances, but also on “a change in administrations.”10 In other words, policy choices are inherent in such decisions, not infiltrated in any sinister sense. Evolving interpretation, at its best, reflects lessons learned through administrative practice—provided always that the new construction remains within the range of interpretive discretion accorded by the statute. Under Chevron, courts remain available to police those statutory boundaries.

This type of structured judicial deference provides a healthy way to understand and facilitate the operation of our administrative state—and to preserve democratic responsiveness to the views of the elected President. Chevron deference also improves the odds that Congress’s broad statutory handiwork, despite its gaps, ambiguities, and internal contradictions or tensions, can be turned into a reasonably coherent and functional operational system by the agency that must make all the pieces fit together. The life of administrative law derives not primarily from judicially ascertained logic, but from experience.11

Against that background, Gonzales and Glen are certainly correct that the referral procedure, though by its nature usable only a few times a year at most, can be a highly useful tool.12 Referral is most significant and effective precisely when the Attorney General steps in to provide authoritative resolution of a messy or complex interpretive issue in a way that will be consistent with wide-horizon policy judgment, as well as with expert understanding of the full regulatory scheme.13 Again, the breadth of that horizon must fall within the range of policy discretion delegated (expressly or implicitly) by the statute. But this leaves much scope for policy-sensitive interpretation in the immigration arena. Despite the copious detail crammed into the pages of the Immigration and Nationality Act (“INA”) over many decades of amendments, the range of interpretive discretion can be extensive, because the INA contains so many poorly crafted and potentially contradictory provisions.14 It is especially appropriate to assign the ultimate decision in this realm to a Cabinet-level official because of the subtle and intricate ways in which immigration decisions sometimes link either to high-level foreign policy concerns or to national security.15

Nonetheless, the Gonzales and Glen Article raises a few questions that deserve closer scrutiny. The first is procedural; the Article’s standards for assessing the adequacy of procedures when referral is granted are not sufficiently demanding.16 I illustrate my critique of these procedural deficiencies in my second major point, which identifies and analyzes the substantive shortcomings of a controversial 2008 referral decision, Matter of Silva-Trevino,17 which was handled under procedures that themselves stirred controversy. To understand those substantive problems, Part III takes a deeper dive into the substantive doctrine at issue before the Attorney General in Silva-Trevino—namely, the “categorical approach” to deciding whether particular offenses amount to “crimes involving moral turpitude” (“CIMT”).18 Better procedure before that decision issued—inviting or at least permitting supplemental briefing by the parties and by amici curiae once the Attorney General had directed referral19—might have led either to a different result or at least to a decision that more fully engaged the true underlying purposes of the doctrine the Attorney General was attempting to alter. Modest but important procedural enhancements would carry significant advantages, both in soundness of result and in acceptance by the public and the bar of the legitimacy of referrals.

II.     Procedures in Referral Cases

Gonzales and Glen spend many pages refuting criticism of the procedures sometimes used in referral cases.20 Much of the criticism centers on two decisions issued late in the George W. Bush administration by Attorney General Michael Mukasey.21 In both decisions, Mukasey adopted interpretations that were seen to narrow the protections or relief possibilities available to individual aliens in removal proceedings. Critics zeroed in on the lack of advance notice of the key issues that were going to claim the close attention of the Attorney General, as well as the failure to invite or even permit additional briefing by either the parties or amici curiae.22

To illustrate both the procedural and substantive issues, I will focus here on one of those Mukasey decisions, Matter of Silva-Trevino. The legal issue in Silva-Trevino was what methodology should be employed in applying the criminal-conviction-based grounds of removal to specific federal or state convictions—in particular, to identify which offenses amount to a CIMT.23[DM1] Gonzales and Glen rise to the defense of the procedures used in that case, and make a strong bid for preserving the flexibility of the Attorney General “to decline cumulative briefing in those cases where he deems it unnecessary.”24

The decisional sequence followed in that case, however, casts doubt on characterizing the procedure the advocates sought as mere “cumulative briefing.” The immigration judge had denied relief, and the respondent appealed to the BIA, arguing simply that the immigration judge had misapplied governing Fifth Circuit precedent.25 The government’s three-paragraph responsive brief before the BIA did not suggest any sort of fundamental challenge to the methodology used in either BIA or Fifth Circuit precedents; it simply asserted the correctness of the immigration judge’s ruling.26 The BIA reversed, in an unpublished decision based on the Fifth Circuit’s doctrine.27 Because the crime was thus not considered a CIMT, Silva-Trevino was potentially eligible for the relief he sought.28 The BIA thus remanded the case for merits consideration of the relief claim.29 Fully a year later, while Silva-Trevino’s counsel was pressing for adjudication in immigration court, he received a terse notice stating that the Attorney General had directed referral of the case from the BIA for final decision.30 Counsel sought information on what issues the Attorney General would be considering, but received no response.31 Three months later, Mukasey issued his lengthy Silva-Trevino decision, making major changes to the Department of Justice’s (“DOJ’s”) previous interpretive approach on the CIMT issue.32

This new and unforeseen ruling galvanized the immigration bar, which sought reconsideration by the Attorney General and an opportunity for supplemental briefing that, unlike the earlier briefs filed with the BIA by the immediate parties, would directly address those issues that were now revealed to be critical to the ultimate legal ruling.33 These were not novel requests for some kind of extraordinary accommodation. As Gonzales and Glen indicate, the DOJ has often permitted or invited supplemental briefing and submissions by amici when a case is deemed sufficiently important to merit consideration on referral to the Attorney General.34 Nonetheless, the Attorney General denied reconsideration.35 Full briefing on the dispositive questions therefore did not occur until the judicial review stage.36

Gonzales and Glen review earlier case law to argue that the Silva-Trevino procedure met the requirements of constitutional due process, whether based on early court precedents addressing these procedural issues in referral cases or a more modern analysis applying the familiar Mathews v. Eldridge calculus.37 Their procedural argument also lays special emphasis on the unlikelihood of “an erroneous deprivation” being visited on the alien as a result of such procedures—both because the individual had an opportunity to be heard and to file briefs before the immigration judge and the BIA, and because further judicial review remains available to catch any errors.

The Article may be right that the underlying procedure, without additional notice or briefing after the Attorney General has accepted referral, would ordinarily meet constitutional due process standards.38 Nonetheless, I would still insist that this part of the Article’s analysis falls short on two counts. First, why make bare satisfaction of constitutional due process the relevant standard? As Alexander Bickel once observed, “to call [an action] constitutional is no more of a compliment than it is to say that it is not intolerable.”39 The Constitution sets absolute minimums that procedures must not fall below. The Due Process Clause, therefore, is not a roving commission for judges to go beyond the minimum and impose their own preferred or optimal procedures.40 But the head of an agency with adjudicative responsibilities does have such a commission—indeed, in my view, a duty to consider providing more than minimally-constitutional procedures—in an effort both to improve decisional outcomes and to assure that the affected parties really have a fair opportunity to be heard on the determinative issues in the specific case. Better procedures can enhance the legitimacy and acceptance of a tribunal’s decisions, even among those who disagree with the ultimate substantive outcome.41

The truly relevant question that I wish the Article had addressed is this: what procedure would best contribute to sound decision-making on those limited occasions when we ask an incredibly busy Cabinet official to devote her scarce time to definitive resolution of difficult issues that combine both law and policy? The fact that the immediate parties to the case may have had an opportunity to file briefs before the immigration judge and BIA, back when no one could have known that the decision would be published as a precedent (much less be issued with the full authority of the Attorney General personally) may meet minimal due process standards. But we should want to equip the Attorney General with a more robust exploration of the real stakes and the full ramifications of choosing one eligible interpretation over another.

Second, the quest to permit more briefing once the case had been referred was not fundamentally concerned with whether Silva-Trevino himself might be subjected to an erroneous deprivation of liberty or property. Upon referral, the case became a broad contest over legal interpretation, and only incidentally a resolution of the specifics of Silva-Trevino’s removal charges. The Attorney General’s acceptance of a referral is rarer than a Supreme Court grant of certiorari.42 Neither procedure is intended—or suitable—for correcting errors in individual cases. These are appellate procedures that are appropriate largely for the resolution of important and recurring legal issues with implications for a wide range of future cases.

As indicated in Part I, Chevron deference is premised on a finding that Congress has not directly spoken to the interpretive issue at hand; the statute is silent or ambiguous.43 In these settings—the kinds of cases where the referral procedure is most likely to be advantageous and efficient—there is no one clearly right answer, though there can be answers that are clearly wrong, i.e., outside the band of discretion delegated by the statute.44 The Attorney General is therefore going to be choosing among alternatives in order to set a legal interpretation that will then govern or affect a very wide range of upcoming decisions. This is clearly a situation where wider input, especially once the specific interpretive issue has become prominent, is most likely to be helpful to a decision-maker. Moreover, announcement of referral is almost certain to trigger amicus interest from organizations represented by skilled and experienced immigration counsel.45 The engagement of counsel with that level of skill and expertise is important because it is far too common in our immigration system for respondents to have unsophisticated representation, if any representation at all, at least during the early stages of removal proceedings. One would think that attorneys general would generally value a more completely developed set of arguments, containing sure-footed reference both to relevant precedents and to details of on-the-ground practice that might be affected by a ruling one way or the other. Such input should facilitate a better-informed decision; it would also enable the Attorney General to anticipate and address a full range of possible objections to the course ultimately chosen.

III.     Substantive Soundness: The Deficiencies of the Attorney General’s Silva-Trevino Decision

A.     The Early Course of the Case

Silva-Trevino is perhaps the most glaring example of a substantively impaired decision that was probably weakened by the DOJ’s resistance to asking for or receiving further briefs and argument before the referral decision was rendered. The underlying facts were these: in 2004 Silva-Trevino, who had been admitted as a lawful permanent resident in 1962, was convicted under a Texas statute making “indecency with a child” a second-degree felony.46 The Texas judge fined him $250, placed him under community supervision for five years, and ordered him to attend counseling sessions.47 Soon thereafter, the Department of Homeland Security (“DHS”) placed him into removal proceedings.48 He was found deportable, but he applied for adjustment of status—a form of relief from removal that potentially could have allowed him to retain or restore his green-card status despite his baseline deportability.49 Such relief was unavailable, however, if his conviction were to be counted as a “crime involving moral turpitude”—a long-standing criterion for removability under U.S. immigration law.50 There have been thousands of immigration decisions adjudicating whether particular offenses under state or federal law count as a CIMT.51

The immigration judge rejected Silva-Trevino’s application for relief, but the BIA reversed, applying a version of the traditional “categorical approach” to the CIMT question, as dictated by the BIA’s understanding of Fifth Circuit case law.52 The categorical approach provides that the moral turpitude inquiry depends on the “statutory crime definition as interpreted by the state’s courts, without regard to the particular circumstances surrounding” the offender’s specific violation.53 If the statute can result in a conviction for behavior that does not involve moral turpitude, then no conviction under the provision will be considered a CIMT, “whatever its actual facts.”54 This more abstract approach saves immigration judges and the BIA from the need to relitigate the facts underlying a conviction, and it also provides greater predictability about outcomes, as case law gradually provides authoritative CIMT characterizations for specific state or federal criminal statutes.55 Because the Texas statute could cover minimally intrusive contact with a minor in settings where the accused was unclear about the age of the victim, the BIA found that a conviction could result from actions that do not involve moral turpitude.56 Therefore it ruled that Silva-Trevino was statutorily eligible for relief.57

B.     The Case on Referral to the Attorney General

Dissatisfaction with the categorical approach had long simmered within some offices in the DOJ, however, because it sometimes lets persons escape removal despite conviction of a highly objectionable crime. Perhaps more to the point, several circuits had adopted varying approaches in applying the categorical approach.58 A complex circuit split generally constitutes a worthwhile setting for the grant of Attorney General referral to provide definitive resolution of the administrative interpretation of the CIMT statute. In principle, such a resolution could command Chevron deference and therefore, under the Supreme Court’s Brand X decision, require even those circuits that had already pronounced on the interpretive issue to yield to the clarified administrative stance.59 Attorney General Mukasey was quite explicit in his Silva-Trevino opinion that he hoped to achieve such uniformity across circuits.60

Mukasey’s opinion described the varying stances taken toward the CIMT inquiry in different circuits.61 Most pertinently, he emphasized that a few circuits had permitted a “modified categorical inquiry” in some circumstances, which allowed the immigration judge to look beyond the statute and into some of the specifics of the particular offense committed by the alien.62 Most courts allowing such a variation permitted the immigration judge to look only to a limited range of outside sources—namely, to the record of conviction, which can include the indictment or information, jury instructions, a plea agreement, and the transcript of a plea colloquy.63 They impose this limitation on outside sources in order to provide a “clear stopping point to” the factual inquiry and thereby avoid “complete re-litigation of past crimes” in immigration court.64

Attorney General Mukasey’s decision rejected sole reliance on the traditional categorical approach.65 He reasoned that it “provides no answer where a statute encompasses both conduct that involves moral turpitude and conduct that does not.”66 This phrasing reveals a remarkable blind spot, because the categorical approach of course does provide an answer: the offense does not count as a CIMT. Obviously the Attorney General was not prepared to accept that result.

Nonetheless, Attorney General Mukasey did not completely reject the categorical approach. The new three-step inquiry announced in his opinion called for starting with the categorical approach; if all offenses under the statute were CIMTs, then immigration judges would efficiently reach the desired conclusion.67 But, if the criminal offense was not categorically a CIMT, then as a second step, the immigration judge would look to the record of conviction to ascertain whether moral turpitude was present—that is, the immigration judge would employ the modified categorical approach.68 Even this, however, is not the end of the inquiry. The opinion further stated: “[I]f the record of conviction does not resolve the inquiry, [the adjudicator should] consider any additional evidence the adjudicator determines is necessary or appropriate to resolve accurately the moral turpitude question.”69 Again the blind spot: the matter is not resolved until every possible step has been taken to see whether a negative result, at least from the standpoint of the alien respondent, can possibly be reached.

The casualness of the Attorney General’s assumption that the statute should be read to authorize locating any evidence of moral turpitude is striking. To be sure, the opinion discussed a few indications in the statutory text that a specialized inquiry to ferret out individual turpitude was intended.70 But that observation was paired with an immediate recognition that other statutory language suggested that the inquiry should be categorical.71 In sum, “the text actually cuts in different directions”72 and does not resolve the issue. For Attorney General Mukasey, this statutory ambiguity was enough for the agency to claim authority to issue a definitive interpretation and to demand Chevron deference.

C.     The Shortcomings of the Decision

Here is where the opinion went astray, and where more ample briefing might have prevented a gap in the reasoning. Chevron deference, for all its prominence, is not the only possible guiding or controlling principle when a statute is ambiguous. As an example, consider retroactive statutes—statutes that attach negative consequences to actions taken before enactment. In the criminal realm, the Ex Post Facto Clause poses a solid constitutional barrier to such retroactivity.73 The Supreme Court has declined, however, to apply a similar restriction, under the doctrinal framework of the Due Process Clause, to statutes imposing civil consequences retroactively.74 Congress generally has broad constitutional authority to adopt such a retrospective statute.75 But, as a counterweight, and in view of the potentially severe consequences for unwitting individuals, the Court has imposed a clear statement rule: statutes will not be interpreted to apply retroactively unless Congress is abundantly clear that such is the intention.76 Ambiguity, in other words, must lead to a legal conclusion that the statute does not apply retroactively, no matter what an administering agency might prefer as a matter of policy.

Similarly, Congress has broad apparent authority under Article III of the Constitution to restrict the appellate jurisdiction of the federal courts and also to provide exceptions to the appellate jurisdiction of the Supreme Court.77 These are potentially sweeping powers, and could raise serious concerns if employed to bar effective remedies for constitutional violations by government officials. Over the past six decades, several bills have been introduced that would have restricted such jurisdiction in response to controversial Supreme Court decisions, and a few enacted federal statutes have seemed to fit that description.78 But our polity has thus far avoided a full confrontation over the exact boundaries of these congressional powers because the Court has adopted another counterweight—a well-established interpretive principle that construes any ambiguity in an ostensible court-stripping statute so as to preserve some avenue to effective court review.79 In INS v. St. Cyr, for example, the Court considered a 1996 statute rather clearly intended by its congressional authors to deny judicial review of removal orders based on criminal convictions.80 The Court had to stretch to find a statutory gap just big enough for it to rule that Congress had not acted with sufficient clarity to bar review via habeas corpus of such a removal order.81 It reached this result even though this judicial construction would give deportation respondents who have criminal convictions more layers of review than other aliens contesting a removal order.82

These kinds of interpretive counterweights make the most sense where Congress’s authority, as a matter of constitutional power, is at its most extensive and where judicially policed limits on that authority are the least available. In such a context, a clear statement rule, which prevents changes from being slipped through with little public or political attention, helps to ensure that when Congress adopts a retroactive civil law or a jurisdiction-stripping provision it is fully cognizant of the true stakes. The rule thus gives opponents of the more severe outcome the best chance to marshal their counterarguments in the political arena.83 It is a device to improve the effectiveness of the political check against borderline or potentially ill-advised (but nonetheless constitutional) laws.

These same considerations apply to the INA’s crime-based removability provisions. For 125 years, the Supreme Court has recognized Congress’s plenary power in the immigration realm.84 Congress has sweeping authority to impose and revise the grounds for removal.85 But the categorical approach has long operated as a kind of implicit clear-statement rule—a modest but useful counterweight.

This kind of counterweight is especially important with regard to crime-based removal grounds, in a way that probably is not intuitive or evident to most casual observers. From one angle of approach, crime-based grounds of removability would seem to be the most easily justified, far less objectionable than removal based on, say, a student visa holder enrolling in classes that come one credit short of a full course load or moonlighting at a pizza parlor without work authorization and thus falling out of status.

But as it happens, the criminal grounds of removability infrequently come into play in cases involving either nonimmigrants or surreptitious border crossers.86 A criminal conviction, to be sure, may be the reason why DHS learned of such a person’s being out-of-status and has prioritized her case for removal. But DHS rarely needs to deploy the crime-based removal ground in order to obtain a removal order in those sorts of cases.87 Proving the more straightforward fact of overstaying or being present without inspection is enough to support a fully effective removal order.88

Crime-based grounds of removal therefore are formally charged, overwhelmingly, in cases that involve a lawful permanent resident (“LPR”). LPRs are people who, at an earlier stage (40 years earlier in Mr. Silva-Trevino’s case), were specifically approved by the United States government for an indefinite stay, in a status that invites sinking roots and establishing community ties founded on permanence. For them, a sweeping exercise of Congress’s plenary power to decree removal potentially carries the most severe impacts, on the respondent alien as well as his or her family, friends, and community.89 In such a situation, a sub-constitutional counterweight is desirable: an implicit clear-statement rule that maximizes the chances that Congress will think more carefully of the individual stakes before making specific offenses a ground for removal of a lawful permanent resident.90

Nothing in the Silva-Trevino opinion acknowledges the uniquely strong impact that capaciously sweeping crime-based grounds for removal have on LPRs. Nor does it recognize, much less engage, the argument that the categorical rule may find justification in broader considerations about appropriate sub-constitutional constraints or counterweights that should operate on broad congressional powers. For good reason, Congress has chosen not to make every criminal conviction the basis for removing LPRs.91 Its dividing lines have often been unclear, requiring interpretation. Using a principle of interpretation that tilts unclear cases toward permitting the LPR to remain makes sense. Such a principle does not preclude Congress from adopting a more severe statute. But if Congress is to do so, it must speak with great clarity. Nor does the nonremoval result necessarily condone crime or devalue deterrence. The criminal law’s penalties will still attach—subjecting the person to the punishments that the state considers sufficient for all other long-time residents in the jurisdiction—that is, the citizens who may be the LPR’s friends, neighbors, or family members.

To be sure, proponents of the more severe regime manifested in the 2008 Silva-Trevino decision by Attorney General Mukasey could probably marshal arguments against treating the categorical approach as a modest palliative for the extensive reach of the plenary power doctrine. But the point is that no such arguments appear in the opinion. Had full briefing been invited or allowed, it seems likely that these important considerations, as well as other detailed arguments cutting against the interpretation ultimately adopted, would have been aired and at least addressed. Conceivably, they could have led to a different result.

In any case, time has not been kind to the 2008 Silva-Trevino decision. The Fifth Circuit overruled Attorney General Mukasey in this specific case, when it reviewed the negative determination that Silva-Trevino received on remand from the Attorney General’s ruling.92 The court found that the Attorney General’s interpretation violated the plain language of the statute.93 Five other federal circuit courts also rejected Attorney General Mukasey’s approach, while two accepted it.94 In the meantime, a cascade of Supreme Court decisions considering analogous issues (interpreting federal statutes that make federal consequences turn on state court convictions meeting certain generic descriptions) likewise steered clear of his methodology.95 Today, a Supreme Court majority appears firmly committed to a strong version of the categorical approach, grudgingly recognizing only two narrow circumstances where some modification of the classic categorical approach is permitted.96 And even in those cases the further inquiry may well be confined to the record of conviction or comparable sources from the original judicial proceeding. Eventually Attorney General Holder bowed to this strong weight of authority and vacated the 2008 Silva-Trevino decision in its entirety in 2015.97

One cannot completely attribute the 2008 Silva-Trevino decision to inadequate procedures, of course, but one would think that this substantive track record—the ultimate decisive rejection of Attorney General Mukasey’s third step in the CIMT inquiry, and a subsequent swing back to a stronger version of the categorical model—would weigh more heavily in the Article’s ultimate assessment of referral procedures. This record should cause Gonzales and Glen to temper their sweeping judgment that allowing the Attorney General to hear a case without supplemental briefing, unless he chooses in an ad hoc process to invite it, “has worked admirably for 75 years.”98

IV.      Conclusion

My strong critique of one particular Attorney General decision should not detract from the overall verdict of Gonzales and Glen on the potential value of the referral procedure. It makes sense to provide for this kind of selective intervention by a Cabinet official bearing broad administrative responsibility in the regulated arena, both to settle open interpretive questions in light of wide-horizon policy judgments and to provide broad and authoritative guidance to executive branch officers. But the Attorney General’s acceptance of any referral case portends an enduring resolution of significant legal questions. The procedure will work better and is more likely to reach sound outcomes if the Attorney General, at the time when referral is accepted or directed, explicitly clarifies the questions to be considered and allows time for supplemental briefing, by both parties and amici, in every such case.

  1. [1]. Alberto R. Gonzales & Patrick Glen, Advancing Executive Branch Immigration Policy Through the Attorney General’s Review Authority, 101 Iowa L. Rev. 841 (2016). Both authors have considerable experience in the Department of Justice, and Dean Gonzales, of course, was the steward of the referral power during his tenure as Attorney General of the United States.

  2. [2]. 8 C.F.R. § 1003.1(h) (2015).

  3. [3]. Gonzales & Glen, supra note 1, at 896–98.

  4. [4]. See Chevron, U.S.A., Inc., v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 842–45 (1984).

  5. [5]. See, e.g., Margaret H. Taylor, Behind the Scenes of St. Cyr and Zadvydas: Making Policy in the Midst of Litigation, 16 Geo. Immigr. L.J. 271, 288 (2002); Justin Chasco, Comment, Judge Alberto Gonzales? The Attorney General’s Power to Overturn Board of Immigration Appeals’ Decisions, 31 S. Ill. U. L.J. 363 (2007); Laura S. Trice, Note, Adjudication by Fiat: The Need for Procedural Safeguards in Attorney General Review of Board of Immigration Appeals Decisions, 85 N.Y.U. L. Rev. 1766 (2010).

  6. [6]. See, e.g., Peter J. Levinson, A Specialized Court for Immigration Hearings and Appeals, 56 Notre Dame Law. 644, 650 (1981); Chasco, supra note 5, at 381–83. See generally Stephen H. Legomsky, Learning to Live with Unequal Justice: Asylum and the Limits of Consistency, 60 Stan. L. Rev. 413, 457–62 (2007).

  7. [7]. See generally Chevron,467 U.S. 837; Abbe R. Gluck, What 30 Years of Chevron Teach Us About the Rest of Statutory Interpretation, 83 Fordham L. Rev. 607 (2014).

  8. [8]. Chevron, 467 U.S. at 842–45.

  9. [9]. Id. at 863–64.

  10. [10]. Nat’l Cable & Telecomms. Ass’n v. Brand X Internet Servs., 545 U.S. 967, 981 (2005).

  11. [11]. My assertion lines up well with the following observation by Cass Sunstein and Adrian Vermeule: “Even in the aftermath of legal realism, some people believe that the interpretation of ambiguities calls for purely legal skills—as it plainly does not.” Cass R. Sunstein & Adrian Vermeule, The Unbearable Rightness of Auer 3 (Harvard Pub. Law, Working Paper No. 16-02, 2016),

  12. [12]. Gonzales & Glen, supra note 1, at 896–98.

  13. [13]. This is often cited as the principal benefit of administrative regimes (and there are many) that place ultimate adjudicative authority in the agency head. See, e.g., Paul R. Verkuil et al., Report for Recommendation 92-7: The Federal Administrative Judiciary, Admin. Conf. of the U.S., Recommendations and Reps. 777, 795–96, 801, 986–96 (1992) (ACUS consultant’s report, describing benefits of head-of-agency review); Russell L. Weaver,Appellate Review in Executive Departments and Agencies, 48 Admin. L. Rev. 251, 260–65 (1996)(describing agencies with appellate review boards whose decisions are subject to discretionary review by agency head). To the extent that the advantages rest on the agency head’s ability to draw on broad experience with the full range of subject-matter regulation, it could be argued that the head-of-agency immigration adjudication authority (i.e., the referral power) should have been transferred to the Secretary of Homeland Security in 2003 when virtually all other domestic immigration management functions, save those carried out by the Executive Office for Immigration Review, were transferred from DOJ to DHS. (The argument would actually support transferring all of EOIR, which includes the immigration judges and the BIA, while of course maintaining or strengthening guarantees of decisional independence for IJs and Board members in resolving individual cases.) I have explored that argument elsewhere, see David A. Martin, Immigration Policy and the Homeland Security Act Reorganization: An Early Agenda for Practical Improvements, 80 Interpreter Releases 601, 614–18 (2003), but will not revisit it here; there is no political appetite today for reopening the issue. A more problematic result of that separation of EOIR from the primary operative agency has been the steadily worsening mismatch of adjudication staffing with enforcement activity. The immigration judge corps has grown only modestly over the past decade while Congress greatly enhanced enforcement funding. See generally Nat’l Research Council, Budgeting for Immigration Enforcement: A Path to Better Performance (Steve Redburn et al., 2011). The DOJ for many years simply did not adequately prioritize the budget increases needed for the IJ corps to grow in a way that keeps pace with caseload. New removal cases are now being calendared in some locations for dates four years later.

  14. [14]. To pick just one small example out of many possible illustrations, Congress has for decades precluded a procedure called “adjustment of status” for persons who worked without authorization or who let their nonimmigrant status lapse, but provided an exception still permitting adjustment for “immediate relatives” of U.S. citizens (meaning spouses, minor unmarried children, and parents). Immigration and Nationality Act, Pub. L. No. 82-414, §245(c)(2), 66 Stat. 163, 217 (1952) (codified as amended at 8 U.S.C. § 1255(c)(2)). In 1996, however, Congress added a new subsection 245(c)(8) that also forbids adjustment of someone who has worked without authorization, but without an exception for immediate relatives. In making this addition, Congress did not disturb the language of that exception in subsection (2). INS decided that the more specific language of subsection (c)(2) would control, thus preserving an immediate relative’s eligibility for adjustment. See Memorandum from Louis D. Crocetti, Assoc. Comm’r, Immigration & Naturalization Serv., to Reg’l Dirs. (Dec. 20, 1996).

  15. [15]. To be clear, most immigration decisions do not carry such high-policy implications. But the subset of decisions that do or might have significance for foreign policy (even in subtle or long-developing ways) is sufficiently important and potentially fast-moving that direct (albeit infrequent) linkage to Cabinet-level resolution, conveyed through a formal and reasoned precedent decision, is valuable. See Arizona v. United States, 132 S. Ct. 2492, 2499 (2012) (“The dynamic nature of relations with other countries requires the Executive Branch to ensure that enforcement policies are consistent with this Nation’s foreign policy....”); Immigration & Naturalization Serv. v. Aguirre-Aguirre, 526 U.S. 415, 425 (1999) (“[W]e have recognized that judicial deference to the Executive Branch is especially appropriate in the immigration context where officials ‘exercise especially sensitive political functions that implicate questions of foreign relations.’”) (quoting Immigration & Naturalization Serv. v. Abudu, 485 U.S. 94, 110 (1988)); David A. Martin, Why Immigration’s Plenary Power Doctrine Endures, 68 Okla. L. Rev. 29, 39–50 (2015) (describing the scope of potential foreign relations linkages).

  16. [16]. See generally Gonzales & Glen, supra note 1, at 896–917.

  17. [17]. Silva-Trevino, 24 I. & N Dec. 687 (Attorney Gen. 2008), vacated, Silva-Trevino, 26 I. & N. Dec. 550 (Attorney Gen. 2015).

  18. [18]. The categorical approach is potentially applicable to many other statutory issues as well, for example in deciding what crimes constitute aggravated felonies or domestic violence offenses, and modifications of that approach in one realm generally would seem to invite or require comparable changes in the others. But the opinion in Silva-Trevino, on its face, was explicitly limited to pronouncing a method for CIMT cases. See generally Silva-Trevino, 24 I. & N. Dec. at 687–90, 698 n.1, 704.

  19. [19]. Referral was directed by Attorney General Gonzales, but the matter was decided and the decision issued by his successor, Michael Mukasey. Silva-Trevino, 24 I. & N. Dec. at 647.

  20. [20]. Gonzales & Glen, supra note 1, at 902–12.

  21. [21]. Id. at 904–06. The two cases are Silva-Trevino, 24 I. & N. Dec. 687, and Matter of Compean, 24 I. & N. Dec. 710 (Attorney Gen. 2009) (revising longstanding standards for assessing claims of ineffective assistance of counsel), vacated, 25 I. & N. Dec. 1 (Attorney Gen. 2009) (by incoming Attorney General Eric Holder).

  22. [22]. As Gonzales and Glen point out, the procedural objections do not so easily fit the Compean decision process, where supplemental briefing was allowed. Gonzales & Glen, supra note 1, at 904–05. Perhaps ironically, Compean drew swift action from the incoming Obama administration to vacate the Mukasey decision—a mere six months after it was issued. Compean, 25 I. & N. Dec. 1 (Attorney Gen. 2009). Despite strong urging from the immigration bar, however, Attorney General Eric Holder did not act to vacate the Silva-Trevino ruling until the final month of his tenure as Attorney General. Silva-Trevino, 26 I. & N. 550 (Attorney Gen. 2015). And he then acted primarily on the basis of numerous case-law developments that had substantially undercut the result and rationale adopted by Attorney General Mukasey, as described in Part III below.

  23. [23]. See, e.g., INA §§ 212(a)(2)(A)(i)(I), 237(a)(2)(A)(i), 8 U.S.C. §§1182(a)(2)(A)(i)(I), 1227(a)(2)(A)(i) (2012)).

  24. [24]. Gonzales & Glen, supra note 1, at 171.

  25. [25]. This account draws from Trice, supra note 5, at 1777–78.

  26. [26]. Id. at 1778.

  27. [27]. Id.

  28. [28]. Id.

  29. [29]. Id.

  30. [30]. Id.

  31. [31]. Id.

  32. [32]. Id.

  33. [33]. Id.

  34. [34]. Gonzales & Glen, supra note 1, at 909 (acknowledging that “additional briefing can benefit a decision”).

  35. [35]. Trice, supra note 5, at 1777–78.

  36. [36]. Id.

  37. [37]. Gonzales & Glen, supra note 1, at 905–12; see also Mathews v. Eldridge, 424 U.S. 319, 335 (1976). The Eldridge calculus calls for a close assessment, in the specific decisional context, of: (1) the individual’s interest; (2) the government’s interest; and (3) the gain to accurate decision-making that can be expected from the exact procedural protection sought.

  38. [38]. Nonetheless, the complete failure of notice in this specific case—even to the deportation respondent—that the attorney general had taken referral in order to provide a comprehensive rethinking of long-established methodology does raise due process concerns.

  39. [39]. Alexander M. Bickel, The Least Dangerous Branch: The Supreme Court at the Bar of Politics 129 (1962) (quoting Charles P. Curtis, A Modern Supreme Court in a Modern World, 4 Vand. L. Rev. 427, 433 (1951)).

  40. [40]. Vt. Yankee Nuclear Power Corp. v. Nat. Res. Def. Council, 435 U.S. 519, 542–49 (1978).

  41. [41]. See generally Laurens Walker et al., The Relation Between Procedural and Distributive Justice, 65 Va. L. Rev. 1401 (1979).

  42. [42]. Trice, supra note 5, at 1767 n.8 counted 17 decisions on referral from 1999–2009, an average of fewer than two cases per year. The Supreme Court typically grants certiorari and hears argument in about 80 cases per year. Frequently Asked Questions, Supreme Court of the United States, (last visited June 21, 2016).

  43. [43]. Chevron, U.S.A., Inc. v. Nat. Res. Def. Council, Inc.,467 U.S. 837, 842–43 (1984).

  44. [44]. See Daniel Kanstroom, The Better Part of Valor: The REAL ID Act, Discretion, and the “Rule” of Immigration Law, 51 N.Y. L. Sch. L. Rev. 161, 167 (2007) (“[T]here is no such thing as a uniquely correct discretionary decision. There may, however, be incorrect discretionary decisions....”).

  45. [45]. The authors seem to complain that the quest to allow such amicus filings materializes only when the attorney general ruling disfavors the individual respondent. Gonzales & Glen, supra note 1, at 912. This may be generally true, but it is not evident why that observation is relevant. Nonetheless, today, immigration cases increasingly attract amicus submissions from organizations, such as the Immigration Reform Law Institute, seeking to support the government position—or indeed to suggest an even more restrictive or enforcement-minded outcome.

  46. [46]. Silva-Trevino, 24 I. & N. Dec. 687, 690 (Attorney Gen. 2008).

  47. [47]. Id. at 691.

  48. [48]. Id.

  49. [49]. Id.

  50. [50]. Id. at 690–92.

  51. [51]. See Dan Kesselbrenner & Lory D. Rosenberg, Immigration Law and Crimes § 6:2 (2011) (summarizing and citing cases).

  52. [52]. Silva-Trevino, 24 I. & N. Dec. at 690–92.

  53. [53]. Id. at 692 (quoting the Board’s opinion).

  54. [54]. Id.

  55. [55]. See Mellouli v. Lynch, 135 S. Ct. 1980, 1986–87 (2015) (discussing the development of the categorical approach); Alina Das,The Immigration Penalties of Criminal Convictions: Resurrecting Categorical Analysis in Immigration Law, 86 N.Y.U. L. Rev. 1669, 1701, 1746 (2011) (discussing the categorical approach); Jennifer Lee Koh, The Whole Better than the Sum: A Case for the Categorical Approach to Determining the Immigration Consequences of Crime, 26 Geo. Immigr. L.J. 257, 295, 307 (2012) (discussing the large numbers of people who rely on this type of case law).

  56. [56]. Silva-Trevino, 24 I. &N. Dec. at 690–92.

  57. [57]. Id. at 692.

  58. [58]. Id. at 693–94 (summarizing the circuit split).

  59. [59]. See Nat’l Cable & Telecomm. Ass’n v. Brand X Internet Servs., 545 U.S. 967, 969 (2005) (“A court’s prior judicial construction of a statute trumps an agency construction otherwise entitled to Chevron deference only if the prior court decision holds that its construction follows from the unambiguous terms of the statute and thus leaves no room for agency discretion. This principle follows from Chevron itself.”).

  60. [60]. Silva-Trevino, 24 I. &N. Dec. at 702.

  61. [61]. Id. at 693–94.

  62. [62]. Id. at 708.

  63. [63]. See Descamps v. United States, 133 S. Ct. 2276, 2283–84 (2013); Rebecca Sharpless, Toward a True Elements Test: Taylor and the Categorical Analysis of Crimes in Immigration Law, 62 U. Miami L. Rev. 979, 983–85 (2008).

  64. [64]. Silva-Trevino, 24 I. & N. Dec. at 702.

  65. [65]. Id.

  66. [66]. Id. at 698.

  67. [67]. Silva-Trevino, 24 I. & N. Dec. at 704.

  68. [68]. Id.

  69. [69]. Id. (emphasis added).

  70. [70]. Id. at 693. In context, the statutory indications are not very persuasive, especially in the light of other textual factors that ultimately led the court of appeals to reject Mukasey’s legal conclusions. Silva-Trevino v. Holder, 742 F.3d 197, 203–04 (5th Cir. 2014).

  71. [71]. Silva-Trevino, 24 I. & N. Dec. at 702.

  72. [72]. Id. at 693.

  73. [73]. U.S. Const. art. I, §§ 9, 10.

  74. [74]. See, e.g., Harisiades v. Shaughnessy, 342 U.S. 580, 591–92 (1952).

  75. [75]. Usery v. Turner Elkhorn Mining Co., 428 U.S. 1, 17–20 (1976) (discussing how an act imposing retrospective civil liability must satisfy only the rational basis test).

  76. [76]. Immigration & Naturalization Serv. v. St. Cyr, 533 U.S. 289, 315–16 (2001);Landgraf v. USI Film Products,511 U.S. 244, 266–73 (1994).

  77. [77]. See U.S. Const. art. III, §§ 1, 2, cl. 2.

  78. [78]. See Erwin Chemerinsky, Constitutional Law: Principles and Policies 149–51 (4th ed. 2011) (describing such bills and statutes since the 1950s).

  79. [79]. See, e.g., Felker v. Turpin, 518 U.S. 651, 659–63 (1996) (discussing Ex parte Yerger, 75 U.S. (8 Wall.) 85 (1869)).

  80. [80]. St. Cyr, 533 U.S. at 308–14.

  81. [81]. Id.

  82. [82]. After St. Cyr, persons with criminal convictions were able to challenge their removal orders on a habeas petition in the district court, retaining the customary second-round appellate review—even though Congress had acted in the same 1996 statute to streamline review by requiring all other challenges to a removal order to proceed directly to the court of appeals on a petition for review. Justice Scalia forcefully pressed this anomaly in his dissent. Id. at 346–47. Congress ultimately responded to St. Cyr by amending the jurisdiction-limiting provisions to make abundantly clear that habeas too would be restricted and that the only review (with very narrow exceptions) was to take place via petition for review in the courts of appeals. But Congress made one other important substantive change that wound up avoiding a full constitutional confrontation with judicial authority. It added INA § 242(a)(2)(D), 8 U.S.C. §1252(a)(2)(D) (2012), which preserves the authority of the court of appeals, notwithstanding other statutory limitations, to consider any “constitutional claims or questions of law” in passing on the petition for review.

  83. [83]. See Landgraf v. USI Film Products, 511 U.S. 244, 272–73 (1994).

  84. [84]. See, e.g., T. Alexander Aleinikoff, David A. Martin, Hiroshi Motomura, & Maryellen Fullerton, Immigration and Citizenship: Process and Policy 162–96 (7th ed. 2012); Martin, supra note 15.

  85. [85]. See, e.g,, Harisiades v. Shaughnessy, 342 U.S. 580 (1952); Fong Yue Ting v. United States, 149 U.S. 698 (1893).

  86. [86]. See David A. Martin, Graduated Application of Constitutional Protections for Aliens: The Real Meaning of Zadvydas v. Davis, 2001 Sup. Ct. Rev. 47, 92–95.

  87. [87]. See id.

  88. [88]. Aleinikoff et al., supra note 84, at 681.

  89. [89]. See Martin, supra note 86, at 92–95.

  90. [90]. There are other categories of cases where the criminal grounds may be decisive though they do not serve as the actual charge on which the notice to appear is based. These, however, primarily involve persons with family ties in the United States that could serve as the basis for adjustment of status, so long as their prior criminal conviction does not run afoul of the inadmissibility grounds. (Silva-Trevino’s case technically involves his claim for that sort of relief through adjustment.) In other related settings, the respondent may be able to use a family relationship or some related finding, such as extreme hardship to a citizen or LPR family member, as the basis for another form of waiver—again so long as his conviction is not considered to fall into a category that bars the waiver. See, e.g., INA § 212(h), 8 U.S.C. §1182(h) (2012) (providing a possible waiver from crime-based inadmissibility, but disqualifying the alien if the conviction was for an offense considered an aggravated felony as defined in INA § 101(a)(43), 8 U.S.C. § 1101(a)(43); the categorical approach is often relevant in deciding whether an offense is an aggravated felony, see Lara-Chacon v. Ashcroft, 345 F.3d 1148 (9th Cir. 2003)). These other types of cases implicating the categorical approach that do not involve someone already possessing LPR status generally do involve, in overwhelming proportions, persons with close family ties to the United States as well as an arguable basis for gaining or retaining permanent residence in connection with those ties.

  91. [91]. A clear example may be found in INA § 237(a)(2)(A)(i), 8 U.S.C. § 1227(a)(2)(A)(i) (2012), which provides that conviction of a CIMT can be a ground for deportation, but not if the person is guilty of only one such offense committed more than five years after admission.

  92. [92]. See generally Silva-Trevino v. Holder, 742 F.3d 197 (5th Cir. 2014).

  93. [93]. Id. at 203–04.

  94. [94]. Silva-Trevino, 26 I. & N. Dec. 550, 553 (Attorney Gen. 2015) (summarizing the judicial treatment of the Mukasey ruling over the previous six years, in Attorney General Holder’s decision vacating his predecessor’s ruling).

  95. [95]. See generally Mellouli v. Lynch, 135 S. Ct. 1980 (2015); Descamps v. United States, 133 S. Ct. 2276 (2013); Moncrieffe v. Holder, 133 S. Ct. 1678 (2013).

  96. [96]. Mathis v. United States, 136 S. Ct. 2243, 2256–57 (2016). The exceptions arise when the crime-related ground of removal is judged “circumstance-specific,” see Nijhawan v. Holder, 557 U.S. 29, 34 (2009), or when the underlying criminal statute is judged “divisible,” see Descamps v. United States, 133 S. Ct. 2276, 2282 (2013).

  97. [97]. Silva-Trevino, 26 I. & N. Dec. 550.

  98. [98]. Gonzales & Glen, supra note 1, at 914.


Warner–Booker Distinguished Professor of International Law Emeritus, University of Virginia. The author served, on leave from the University, as General Counsel of the Immigration and Naturalization Service from 1995 to 1998 and as Principal Deputy General Counsel of the Department of Homeland Security, 2009 through 2010.

The views expressed here are the author’s own and do not represent the views of his former agencies or the U.S. government.