Sunday, November 8, 2015

Torres v. Lynch - Probably Not the Case on Chevron Deference to Agency Interpretations of Criminal Law We've Been Waiting For

In previewing Torres v. Lynch, I asked whether Torres was the case on Chevron deference to agency interpretations of criminal law we've all been waiting for.  Torres, I explained, was a case where the agency interpretation of a defined term in immigration law, "aggravated felony," had consequences for the meaning of several immigration-related crimes that use that term, where the petitioner and amici argued vigorously that the agency should therefore not receive Chevron deference, and where the government, at least in the alternative, seeks Chevron deference should it lose at Step One.  I noted that Justice Scalia (in a statement respecting a denial of cert joined by Justice Thomas) and Judge Sutton have recently raised the issue of Chevron deference to agency interpretations of regulatory provisions that have criminal-law effect, that the statute in Torres likely was ambiguous, and that the Court would likely have to decide whether Chevron applied, absent severing the civil applications of the term at issue from the criminal applications.  (Severing in the sense that they could have different meanings, despite sharing the same definition - a move which has some precedent in the Court's administrative law cases.)

It would appear from the transcript, however, that the Court has absolutely no interest in deciding whether Chevron applies to agency interpretations of criminal law in this case.* There were just two questions about deference during the entire argument, both came from Justice Ginsburg, and neither went to Chevron  deference to interpretations of criminal law.  The first, put to Torres's counsel at the end of his argument, was "What about the -- the respect that we owe to the BIA for their interpretation?"  Torres's counsel replied that the rule of lenity would foreclose Chevron deference, but this didn't elicit any interest in the form of a follow-up.  Later, when the assistant to the SG, Elaine Goldenberg, raised deference at the back end of her argument, Justice Ginsburg asked about the fact that the Board of Immigration Appeals has changed positions on this issue.  Goldenberg replied that the Board had had good reasons for changing its position, and the Chief Justice interjected that in Chevron itself the EPA had changed positions.  There too, nothing was said by the Justices about deference to the Board's (collateral) interpretation of criminal law.

* For a more general recap, see Steve Vladeck's recap for Scotusblog, with which I generally agree, though I think he underestimates support for the petitioner, and which argues that a majority of the Court is inclined to vote for the government and hold that convictions for state offenses "described in" federal statutes include offenses that lack those statutes' jurisdictional elements. Listening to the argument audio, I hear at least one solid vote for Torres - Justice Sotomayor - and four more possible votes for Torres - Justices Scalia (with whom Justice Thomas could follow), Breyer, Kagan, and Ginsburg, though Kagan and Ginbsurg seem to be leaning in favor of the government. 

What explains this lack of interest in a hot and potentially outcome-determinative issue?  Perhaps the Justices (except Justice Ginsburg, perhaps) think the statute here is unambiguous one way or another, making Chevron's applicability irrelevant.  The government's substantial purposive arguments notwithstanding though, what could be more ambiguous than whether "an offense described in [a federal statute containing a jurisdictional element] . . . whether in violation of Federal or State law" necessarily contains the federal statute's jurisdictional element?  The meaning of "described in" just isn't clear. (See endnote.**)

A clue, I think, is contained in Torres's reply brief, which points out that in the Court's half-dozen engagements with the provision at issue here -- a list of offenses, housed under the heading of "aggravated felonies," that subject aliens to various immigration consequences, including deportation -- the Court has never applied or even mentioned Chevron.  It's true that in the cases he cites, plus one he doesn't mention, the government declined to request Chevron deference in each case but one, usually because the Board of Immigration Appeals hadn't rendered a precedential decision on the questions at issue or had adopted a posture of conforming its decisions to local circuit precedent.  However, as the government did point out in its briefs in one such case, where the Chevron framework applies, the Court has held it will ordinarily remand to the agency if it finds ambiguity.  Yet the Court has never inquired, in any of these cases, into whether or not there was ambiguity, and has decided some of them without remanding in cases where the statutory text at issue was anything but clear.

Perhaps the Court has always been concerned about the collateral criminal consequences of deferring to the Board's interpretation of the aggravated felony definition, or at least concerned enough to avoid deciding whether the Board gets deference so long as the Court could arrive at a tolerably clear interpretation of the definition on its own.  That's a possibility.  But in both of the cases where the government has sought Chevron deference to the Board's interpretation of the definition - this one and Nijhawan v. Holder - its immigrant-opponents have argued that the Board can't get Chevron deference to interpretations of terms that appear in criminal law, and in both cases no one on the Court indicated any interest in that claim at oral argument.  (To the contrary, Justice Scalia, who's been calling for a criminal-law exception to Chevron since 1990, suggested Chevron should apply in the Nijhawan argument.)  Nor does the unanimous opinion in Nijhawan say anything about it.

A more plausible explanation, I think, is that the Court just doesn't think the Board has sufficient expertise in criminal law to warrant deference to its interpretations of generic offenses and cross-references to federal criminal statutes.  On the other hand, the Court  has been fairly consistent about giving the Board deference on "pure" questions of immigration law, probably because, on this theory, the Court does think the Board has sufficient expertise on those sorts of questions to warrant deference.

Of course, expertise is not an official part of the Court's deference doctrine, except (a) in Skidmore, where it goes to the "weight" an agency position gets, and (b), under Barnhart, in deciding whether an agency gets Chevron deference if an agency doesn't satisfy the Mead safe harbors (i.e., if its interpretation isn't set forth in a notice-and-comment rulemaking or formal adjudication).  Neither use of expertise could explain the Court's reliance on it in this context; the Board engages in formal adjudication, which takes it out of both Barnhart and Skidmore.*  

*The courts of appeals generally only give Skidmore deference to the Board's unpublished opinions, but the relevant opinions in Nijhawan and Torres are published, and in the other aggravated-felony cases, where there were no controlling published opinions, the Court could have remanded for the Board to write one.

However, as Bill Eskridge and Lauren Baer demonstrated seven years ago in a landmark empirical study of the Court's deference cases, the Court's deference doctrines are woefully inaccurate descriptors of the Court's deference practices.  In practice, the Court defers to agencies in something like half the cases its doctrines say it should, and in practice, Eskridge and Baer say, agency win rate (if not what, if any, deference doctrine is used by the Court in the first place) appears to have a lot to with agency expertise.  Eskridge and Baer also called on the Court to turn its tacit reliance on expertise into official doctrine; seven years later, that hasn't happened.  To the contrary, the Court has only become more emphatic about Chevron's inflexibility and scope (King v. Burwell's recognition of the major-questions doctrine notwithstanding).  Consequently (among many other consequences), the courts of appeals uniformly defer to the Board on the meaning of generic crimes and criminal-law references in immigration law, even as the Court never has for reasons which go unexplained in its opinions.

If the Court really is tacitly denying deference in large swaths of cases, like the cases on the meaning of the aggravated-felony definition, because it believes the relevant agencies lack expertise on the issues those cases present, two questions come to mind.  One, is a lack of expertise really a good reason to deny deference?  Two, if it is, shouldn't the Court say so so that lower courts can deny deference for lack of expertise too?

Briefly, on the first question, I don't think a lack of expertise is a good reason to deny an agency Chevron deference.  To the extent, at least, that a statute is ambiguous or vague enough as to cause serious doubt that Congress ever resolved or thought about a given question (which may not be the case of every ambiguity), the Court errs when it searches for a legislative intention or textual meaning that isn't really there.  And, if Congress has left some policy choice open, that choice should usually be made by an agency, not a court, regardless of the agency's expertise on that particular issue.  An agency, at least, has some political accountability, both to the President and to Congress, and can more openly engage in the policy reasoning that an unmade policy choice demands.  Besides, expertise can often be completely besides the point.  

For example, take Torres itself.  If in Torres the statute really is hopelessly ambiguous on whether "an offense described in Federal Statute X, which contains a jurisdictional element" includes state offenses that lack the jurisdictional element, what expertise is really needed to decide what the law should be (rather than what it is) on that score?  The Board, being comprised of political appointees by the Attorney General and her recent predecessors, can take note of the President's preference for more/less deportation of criminal aliens and simply decide the issue that way.  The Board can also simply decide whether it deems state arson offenses that lack jurisdictional elements worthy of deportation, as Congress itself could have, but didn't.  Either sort of decision, while inexpert, is much more legitimate than (a) the Court straining to find statutory meaning that isn't there, or (b) the Court instantiating its own policy preferences on deportation of criminal aliens, whether openly or in the guise of divining the fixed meaning of the statute.

As to whether the Court should be more open about its (possibly) expertise-based deference regime, I don't think so.  In the first place, I don't think a lack of expertise is a good reason for denying deference, so I wouldn't want it to spread.  But leaving that aside, the Court's silence on the reasons it denies deference may, like Chevron itself, be based on a frank and wise assessment of institutional competence.  It is all very well for the exquisitely gifted and exquisitely well-briefed nine lawyers on the Court to deny inexpert agencies deference in a handful of cases a year.  It would be quite another thing to license the entire federal judiciary to deny inexpert agencies deference in the thousands of cases a year where deference is an option.  Making that move might cause lower courts to substitute their less expert and less informed judgment for that of agencies, and would make litigation over agency-interpreted statutes hopelessly unpredictable, as expertise is very much in the eye of the beholder.  It would also give lower courts that just don't like a particular agency interpretation a ready-made excuse to not defer to it, at least in any case where they could plausibly accuse an agency of lacking expertise.  It seems, then, all for the best that the Court has declined Eskridge and Baer's invitation to raise its deference practices to the status of vertically binding deference doctrine.
** One could argue that "an offense described in Statute Y . . . whether in violation of Federal or State law" unambiguously picks up all the elements of Statute Y, but that reading threatens to make a null set of half of "whether in violation of Federal or State law."  Alternatively, one could argue that given the categorical approach (under which, if Torres is right, immigration judges would have to look to whether a state statute contained a federal jurisdictional element, not just to whether an alien's state offense happened to involve conduct which would satisfy that element), given the "whether in violation of Federal or State law" language, and given the fact that state statutes will not contain federal jurisdictional elements, it just has to be the case that "described in" excludes jurisdictional elements. 

I don't think this quite works.  Not only does the "whether in violation of Federal or State law" not unambiguously require that some state-law violations of the described-in offense actually exist, the categorical approach is itself not an unambiguous mandate.  In fact, the Court has occasionally deviated from the categorical approach in interpreting the very statute at issue in Torres, and could do so again, as Justice Breyer suggested at oral argument.  Were it to do so here, the argument that "described in" has to exclude jurisdictional elements unravels; immigration courts could look for jurisdictional facts in state offenses that matched the federal jurisdictional elements.

Saturday, November 7, 2015

Lockhart v. United States - Argument Recap, and Ruminations on Canon Entrepeneurship, Legal Indeterminacy, and Legislative History

The argument in Lockhart was one of the more fun and rich arguments on statutory interpretation I've read in a while.  Even if you have no interest in the issue in the case (what kinds of prior convictions for sexual abuse increase mandatory minimums for child pornography), you should read this transcript if you care about current trends in statutory interpretation.  With that being said, I'll offer some post-argument predictions, and some comments on the broader themes of the argument.

As readers of this blog will recall, Lockhart's about whether the phrase "involving a minor or ward" modifies only "abusive sexual conduct" in the list "aggravated sexual abuse, sexual abuse, or abusive sexual conduct involving a minor or ward" (a list of sentence-enhancing priors for child pornography sentencing purposes), or modifies all three types of offenses.  The government, awkwardly, reads sexual abuse and abusive sexual conduct as synonyms, with the result that "involving a minor or ward" merely modifies one of two synonymous phrases and thus does no meaningful work.  As far as I can tell, not a single Justice bought this reading of the statute.  Insofar as some Justices were inclined to support the government's bottom line - that "involving a minor or ward" only modifies "abusive sexual conduct" - they clearly intend to get there by reading "abusive sexual conduct" to mean something less than "sexual abuse."  Having made that move, the statute could then plausibly be read to require child victims only for lesser abusive-conduct offenses, but not for more serious sexual-abuse offenses.

How though, one might ask, would one distinguish sexual abuse from abusive sexual conduct?  Isn't abusive sexual conduct sexual abuse by its terms?  One solution, which Justice Alito (and to a lesser extent the Chief Justice) pursued throughout argument with great persistence, is that substantive federal offenses use very similar terms and give them different definitions.  Sections 2241 and 2242 of Title 18 define aggravated sexual abuse and sexual abuse, respectively, to involve abusive sexual acts (which Section 2246 defines to require, at the least, unmediated contact with genitalia), while Section 2244 of Title 18, in tandem with Section 2246, defines "abusive sexual contact" to require abusive contact with clothed genitalia, or various erogenous organs.  Couldn't these definitions explain the elusive distinction between sexual abuse and abusive sexual conduct?

As Ann McConnell, assistant to the SG, forthrightly conceded in fighting off this friendly attempt to rescue the government's case, the federal definitions of sexual abuse and abusive sexual contact probably aren't much help here.  One problem, of course, is that Congress didn't use the phrase "abusive sexual contact" in the child-pornography sentencing provision; it wrote "abusive sexual conduct."  Another problem is that if Congress had really meant to say that state convictions for sexual abuse and abusive sexual contact as defined in federal law were priors, it could have expressly incorporated those definitions - as it in fact did in another recidivist sentencing provision.  

These difficulties notwithstanding, Justice Alito wasn't taking no for an answer.  When McConnell said the government didn't think the child pornography sentencing provision was tacitly referencing the definitions of sexual abuse and abusive sexual contact because it could have explicitly referenced them, Alito sarcastically asked whether abusive sexual conduct and sexual abuse were just "catchy phrases that came to [Congress's] mind," and suggested that while Congress gets "a 'D' for their drafting of this statute," he was nevertheless pretty sure that Congress was thinking of its definition of abusive sexual contact when it wrote about state convictions relating to abusive sexual conduct.

Outside, though, of Alito and the Chief Justice, who seem fairly bent on saving the government from the infirmities of its position, it's difficult to find support in this transcript for the government.  Going down the list, in no particular order:

Justice Kennedy asked a series of noncommittal, narrow questions that eventually centered on the role of lenity in the case, which isn't a great sign for the government.  It should be noted, though, that his questions doubted that lenity should be stronger as to sentencing provisions, so perhaps his concern was merely with assuring himself that only normal-strength lenity applied.

Justice Breyer initially expressed some interest in Alito's theory.  But when it was the government's turn to argue, Breyer challenged McConnell to explain why the statute wasn't at least ambiguous.  After McConnell gave her reasons, he then suggested that he found all of them wanting, and asked why, if the parties' textual arguments were in equipoise, the legislative history of the statute, which described the various sexual abuse priors as child abuse offenses, shouldn't control.  

Justice Scalia entered argument with his mind clearly made up for Lockhart, seemingly taking the view that, while the ambiguities in the statute favored Lockhart on balance, the statute was ambiguous enough that lenity resolved the case in Lockhart's favor.  

Justice Kagan, in the beginning of argument, suggested that abusive sexual conduct possibly covered some set of lesser sexual-abuse offenses (like indecent exposure to minors), and that on that reading, it would make sense if the least severe set of offenses only counted as priors if they involved child victims.  But at the end of McConnell's argument, she stated pretty flatly that in her view, Lockhart's preferred canon for disposing of the case, the newly minted series-qualifier canon (more on this below), on which modifiers modify an entire list in certain circumstances, likely "trumped" the government's preferred canon, the last antecedent rule, under which a modifier modifies the last possible antecedent only.  

Justice Ginsburg didn't say much at argument, but asked the government a few questions - albeit not especially pointed ones - about a letter by an acting assistant attorney general, written a couple years after the sentencing provision at issue in the case was enacted, that appears to have embraced Lockhart's reading of the provision.

Justice Sotomayor asked just two questions.  One was skeptical of the government's view that "abusive sexual conduct involving a minor or ward" may have been added to the statute simply to clarify that statutory rape was a prior; she asked why, if that was the case, Congress didn't just write "sexual conduct involving a minor or ward."  The other question, put on rebuttal to Lockhart's federal public defender, Edward Zas, was why Congress would have made federal sexual abuse crimes enhancing priors irrespective of victim age, but not done the same with state sexual abuse crimes.

Justice Thomas, of course, asked no questions.

Prediction: I think Scalia, Breyer and Kagan to a lesser extent are pretty unavailable votes to the government.  It is barely possible that the government could eke out a narrow majority comprised of Alito, Roberts, and at least three of Sotomayor, Ginsburg, Thomas and Kennedy, but that would require a lot of breaks and a fairly improbable alignment.  It's hard to imagine Ginsburg or Sotomayor providing the fifth vote to affirm Lockhart's sentence while Scalia votes for Lockhart on the basis of lenity.  


Canon entrepeneurship: This case has been framed by the parties as a battle of dueling canons - the rule of the last antecedent, and the series-qualifier rule.  The series-qualifier rule, though, was first discovered by name, at least, in Justice Scalia's 2012 book, Reading Law.  And as Neal Goldfarb points out at his blog, LawnLinguistics, while the series-qualifier rule has some antecedents in caselaw, it was never stated in Scalia's precise formulation, or anything too close to it, until Scalia's formulation.  Particularly, Goldfarb says that the closest thing to series-qualifier in the Court's cases is a fairly distant relative, namely a kind of non-canon that says modifiers modify all their possible antecedents when it makes sense for them to do so, while Scalia's formulation requires global modification in the case of any "straightforward, parallel construction" involving all the verbs or nouns in a list.  Moreover, the somewhat closer cases in the lower federal courts and state courts that Scalia cites never applied series-qualifier or antecedents thereof, prior to Reading Law, to postpositive modifiers (i.e., modifiers that follow the nouns they modified) - the very kind of modifier at issue here.  

But even though series-qualifier may be Scalia's extremely recent invention, and has yet to be mentioned by the Court in a case since Reading Law was published, this didn't stop the government from beginning its argument by framing the case as a battle between "two competing canons of statutory interpretation," or acknowledging at argument that something called "the series-qualifier canon" exists and arguing that it didn't apply.  Nor did it stop Justice Kagan from holding forth on the interplay between "that series-qualifier rule" and the rule of the last antecedent, reciting the elements of the rule as Scalia formulated them, and suggesting that Lockhart's a case where Scalia's canon "trumps" one of the oldest canons around.  Nor did it cause anyone to question Lockhart's counsel on the provenance of a canon for which his brief largely cites to Scalia's book.

What's going on here?  One possibility is that the SG and Justices assume Scalia's book accurately describes the canons and the cases it cites, even though it's been famously pilloried as inaccurate on just these points by Judge Posner.  Another possibility is that the Justices, and especially the SG, find it awkward to question the accuracy of a sitting Justice's scholarship.  I suspect both of these explanations have some merit.  But the most interesting possibility is that the Justices don't much care about the provenance or age of a linguistic canon, so long as it makes a certain kind of sense.  

If that's true, and I think it is, two things follow.  One, that the Court doesn't care whether Congress drafts against the background of a canon or even has advance notice of one, but rather believes that a canon is valuable insofar as it accurately describes congressional linguistic practices and congressional intent (or "objectified" intent, if one's a textualist of a certain stripe).  Two, the Court's apparent openness to new canons, particularly countercanons to existing ones, suggests that the general academic posture to canons, which assumes that they're a fixed set of hidebound hoary maxims that misread congressional intent and criticizes them on that basis, is mistaken.  Instead, we should be talking about how to improve the canons - or, even if one believes that even a perfected set of canons will generate error (which of course it would), debating whether judges would really get cases right more often if they engaged in free-form contextualism.  Unfortunately, while scholarship of that kind could be quite influential, there isn't much of a market for articles on specific canons, especially linguistic ones, which has left courts with virtually no guidance as they formulate rules that purport to reflect legislatures' linguistic practices.

Legal indeterminacy:  One underappreciated and subtle difference between textualism and intentionalism, which this argument brought out, is that textualism is, at least in theory, far more open to the possibility that a statute's meaning is indeterminate.  Scalia and Breyer's differing approaches to the difficult textual ambiguity Lockhart presents nicely illustrate this.  Here is Scalia introducing the rule of lenity into the argument: "we've been discussing these dueling canons and so forth.  My goodness, I have no -- I have no assurance what the right answer is.  But I know that somebody could read this and think it means what the petitioner says it means.  And if that's the case, it seems to me the rule of lenity comes into play." 

Here, it appears to me, Scalia is suggesting that this is possibly a case where statutory meaning is indeterminate.  Now, one might think of lenity as just another canon that determines statutory meaning.  But I think that's wrong.  To rule for a defendant on lenity isn't to make a claim about statutory meaning at all.  It's to say that, in cases of irresoluble ambiguity, a default rule favors criminal defendants, not because anyone thinks that Congress necessarily enacted or intended the defendant-favorable rule, but because of concerns about notice and judicial criminal-lawmaking.  (At most, lenity constructs meaning, but plays no role in interpreting it.)  Because irresolubly ambiguous and vague texts exist, textualism is committed to the view that irresolubly ambiguous and vague statutes exist - which is a part of why, I think, ambiguity-resolving rules like Chevron and lenity have such appeal to Scalia.

On the other hand, Justice Breyer, having found the textual arguments in the case at "equipoise," argued that "before turning to the rule of lenity . . . the legislative history helps," and that the legislative history ultimately made manifest Congress's intentions and thus the meaning of the statute.  For Breyer is not so quick to think that statutory meaning runs out.  As an intentionalist, he believes that even if a statute is textually ambiguous on a question, Congress may nonetheless have had a definite intention on that question, and that that intention is law.  Of course, intentionalists, Breyer included, do not believe that a statute always has some singular intention behind it on every litigable question under the statute.  Chevron, a deeply intentionalist opinion by a committed intentionalist, Justice Stevens, acknowledges that there are cases where Congress lacks "an intention on [a] precise question at issue."  But in theory, those cases are a subset of the cases where textualists find irresoluble ambiguity, which is why, in part, Justice Breyer has never been able to make peace with Chevron as it's commonly understood today - to trigger deference whenever a statute is textually ambiguous.*

* Which isn't at all how it's explained in the Chevron opinion itself.  As Frederick Liu has recently pointed out, in addition to the language I quoted above from footnote 9, Stevens cites a passage from Roscoe Pound in footnote 10, in which Pound discusses "cases in respect to which the lawmaker had no intention."  These, and only these, were the cases where Stevens appears to have envisaged Chevron deference kicking in. 

Continuing debates over legislative history: Reliance on legislative history is at its lowest ebb on the Court right now in nearly a century, but I believe it will make a comeback, because the arguments for its use are getting smarter.  Defenders of legislative history used to conceptualize it as a kind of revelatory gloss on what a statute meant, which opened them up to arguments that intentionalists were allowing Congress to delegate the details of statutory meaning to single members or committee staff, in violation of the non-delegation doctrine and the requirements of bicameralism and presentment.  Today, intentionalists tend to justify legislative history as a synopsis of the legislator/legislative staff-created materials on which legislative counsel-penned statutory text is based.  What is the non-delegation critique of that story?  This new debate played out in the Lockhart argument, as Breyer, the Court's beleagured defender of legislative history, took a minute to defend his reliance on committee materials before and after getting around to asking a specific question:

the report, the way these are actually written is that a general idea is given to a person whose job it is to draft.  And that drafter tries to put in words the general  intention that's been described, and the report is used to describe what the general intention was...

So I read [the committee report] and said, short, but clear. And that's what the drafter would have been looking at when -- the first one, anyway -- working with the staff of the committee when trying to translate general intentions of senators and representatives into actual language. And I think it's not contrary to popular belief to say that senators and representatives do hire staff to do such things and do not sit there with pen and pencil thinking, where does the "or" go?
After Justice Breyer asked this question, Scalia asked the SG's assistant whether she thought "Congress can leave it to its staff to decide what a statute means . . . Isn't legislative power nondelegable?"  This non-delegation critique of Breyer's argument, though, seems like a non sequitur; at what stage in his account of legislation has Congress "le[ft] it to its staff to decide what a statute means"?  What Breyer says it leaves to staff is statutory text (which is undeniable), and if Scalia thinks that violates the non-delegation doctrine, it's hard to see why the response would be an interpretive approach that solely looks to the product of the unlawful delegation.

Monday, November 2, 2015

Lockhart Solved

Tomorrow the Court hears argument, in United States v. Lockhart, on whether the phrase "involving a minor or ward," in a statute that refers to "a prior conviction . . . under the laws of any State relating to aggravated sexual abuse, sexual abuse, or abusive sexual conduct involving a minor or ward," only modifies "abusive sexual conduct" or aggravated sexual abuse and sexual abuse as well.  As a matter of grammar, the case could go either way.  As a matter of statutory interpretation, the rule of the last antecedent says "involving a minor or ward" only modifies the last "antecedent" (really, last noun), absent contextual indicia to the contrary.  One such contextual indicium, explored in my last two posts on this case, is that where the government concedes that abusive sexual conduct and sexual abuse mean about the same thing, it wouldn't have made much sense for Congress to use "involving a minor or ward" to modify one and not the other.  This is the kind of context-specific inference of legislative intent people usually use to evade the last antecedent rule.

It turns out, however, that in Lockhart there's a quasi-grammatical (really, stylistic) argument that knocks out the rule of the last antecedent.  As a general rule of style, when people write out lists of alternatives, they usually insert an "or" before the last item in the list - A, B, C, D, or E - which serves to indicate that the entire list is alternative.  That is, "you can buy A, B, C, D or E" means "you can buy A or B or C or D or E."  Occasionally, people will use "or" between each item on a list of alternatives.  What one never sees is a hybrid between the two, such as "A, B or C or D or E."

However, as an astute commenter on Neal Goldfarb's excellent blog on law and linguistics points out, the statute in Lockhart states a list in just that odd form: A, B, or C, or D, or E.  It says that one gets an enhanced sentence for a child pornography conviction if one has a prior conviction under state law "under the laws of any State relating to aggravated sexual abuse, sexual abuse, or abusive sexual conduct involving a minor or ward, or the production, possession, receipt, mailing, sale, distribution, shipment, or transportation of child pornography, or sex trafficking of children."  To simplify, that's "aggravated sexual abuse, sexual abuse, or abusive sexual conduct involving a minor or ward, or [child pornography offenses], or sex trafficking of children."  Why the "or" before "abusive sexual conduct," but not before "sexual abuse"?  And why the "or" before the child pornography offenses?   Why not simply "aggravated sexual abuse, sexual abuse, abusive sexual conduct involving a minor or ward, [child pornography offenses], or sex trafficking of children"?  

The only explanation, besides positing that Congress just chose to write a list in an extremely eccentric way, is that the first three offenses in the list - aggravated sexual abuse, sexual abuse, and abusive sexual conduct - are a sub-list of offenses "involving a minor or ward" within the larger list of enhancing priors.  Once the first three offenses are read as a sub-list of related offenses, the use of an "or" at the end of that list makes sense.  Likewise, "the production . . . or transportation of child pornography" is a sub-list of pornography offenses within the larger list of enhancing priors, which explains why it seems perfectly natural for "transportation of child pornography," the last item on that list, to both be preceded and followed by an "or."  If the first three offenses are a sub-list of offenses that each involve minors or wards, the "or" before "abusive sexual conduct" wasn't just permissible, but necessary.  "Under the laws of any State relating to aggravated sexual abuse, sexual abuse, abusive sexual conduct involving a minor or ward, child pornography offenses..." could only be read to require minor/ward involvement as to abusive sexual conduct alone.

An argument of this kind usually won't defeat the last antecedent rule, because the last antecedent rule is usually invoked to limit a modifier to modifying the last item on a list, not the third-to-last item on the list, as is the case here.  That is, the statutes where the rule is applied typically go "A, B, C, D, or E that Q," not "A, B, or C that Q, or D or E."  However, in cases where a modifier shows up in a middle of a list, as it does here, the presence or absence of an "or" before the last antecedent should usually (always?) resolve whether the modifier modifies the last possible antecedent or the terms that precede it.

Torres v. Lynch - The Case on Chevron Deference to Agency Interpretations of Criminal Law We've All Been Waiting For?

Torres v. Lynch is a case, on its face, about a deliciously narrow question of immigration law that turns on a peculiar kind of cross-reference.  It's also a case that may turn on whether administrative agencies get Chevron deference to interpretations of regulatory provisions that collaterally interpret criminal law.  

A.  What does an "offense described in" a federal statute mean?

First, some comments on the immigration law.  In the Immigration and Nationality Act, an "aggravated felony" is a kind of federal or state conviction that subjects an alien to a host of serious immigration consequences - deportability, ineligibility for discretionary cancellation of deportation, and ineligibility for asylum, among other things.  (Aggravated felony convictions also have certain criminal law consequences, discussed in Part B below.)  The term is defined, exhaustively, though none too perspicuously, in a 21-subsection subsection of the Act (see subsection (a)(43)), to include a number of crimes, which Congress described in three ways.  First, there are a number of generically described offenses - for example, murder and rape.  Second, there are several offenses, or more accurately, kinds of offenses, "as defined in" some federal statute - for example, "a crime of violence, as defined in 18 U.S.C. 16."   Third, there are a number of offenses - by my count 55 - "described in" one or another federal criminal statute.  This list is followed by a proviso which states that the term "aggravated felony" applies "to an offense described in this paragraph whether in violation of Federal or State law."  The question in Torres is, if someone is convicted of a state offense that includes every element of one of the listed "described in" federal offenses except the jurisdictional element of the federal offense, has he committed the offense "described in" the federal statute?

Getting down to details, 8 U.S.C. 1101(a)(43)(E)(i) says that an aggravated felony includes "an offense described in" 18 U.S.C. 844(i).  844(i), a sentencing provision, provides the sentence for "[w]hoever maliciously damages or destroys, or attempts to damage or destroy, by means of fire or an explosive, any building, vehicle, or other real or personal property used in interstate or foreign commerce or in any activity affecting interstate or foreign commerce."  The petitioner in this case violated a New York arson law that proscribes intentionally damaging, by means of fire or explosion, a building or motor vehicle. Having done so, he committed, element for element, the offense "described in" 844(i) - except that the New York statute, unsurprisingly, doesn't require the building or motor vehicle an arsonist destroys to have been used in interstate or foreign commerce.  So the petitioner wasn't convicted of an offense that requires all the elements 844(i) requires.  Nonetheless, did he commit the "offense described in" 844(i)?  Or did he come one element short of committing the offense described in 844(i)?  The Board of Immigration Appeals, affirming an immigration judge's order deporting Torres, said the former, and they typically get Chevron deference to their interpretations of the immigration laws.  But absent some pretty good reason to think otherwise, the answer almost has to be the latter.  After all, what sense does it make to say that the offense described in 844(i) is arson without the jurisdictional element, when 844(i) plainly describes a kind of arson with the jurisdictional element?

An initial reaction one might have in favor of the Board's interpretation is that "an offense described in . . . 844(i)" suggests something a little different from an offense that perfectly matches 844(i).  The problem with that reaction is that some language like "described in" had to modify 844(i) and the many other enumerated described-in offenses in order to clarify that Congress wasn't just picking up the federal offenses themselves, but also meant matching state offenses (bracketing for the moment how closely they have to match).  Perhaps "described in" solely does the work of making that much clear.  Another initial reaction one might have is that, if "an offense described in [a federal criminal statute]" picks up matching state offenses, it has to exclude the jurisdictional elements from the federal statute, or it won't pick up any state offenses.  But it turns out that, 844(i) notwithstanding, a great many of the 53 other "described in" offenses lack jurisdictional elements.  So reading "described in" to pick them up will still leave a fair amount of "described in" state offense coverage, though virtually no state arson coverage.

As careful readers will have anticipated, the government says that there's an important difference between the listed offenses "as defined in" federal statutes and the listed offenses "described in" federal statutes.  "Defined in," they say, suggests a tighter fit; "described in," a looser fit.  That argument has persuaded several circuits.  But it's a complete loser.  The aggravated felony definition does use "defined in" in several spots instead of "described in," and there is a reason for the different usages.  But it's not the one the government thinks.  When Congress used "defined in," it did so for the simple reason that it was using a defined term, either as part of a description of a crime, or where Congress has a defined term elsewhere in the U.S. Code that describes a category of crimes, and wanted to reference its definition.  See, e.g., 8 U.S.C. 1101(a)(43)(B) ("illicit trafficking in a controlled substance (as defined in section 802 of Title 21), including a drug trafficking crime (as defined in section 924(c) of Title 18).").  Not once in the aggravated felony definition does Congress simply refer to "an offense as defined in" some provision, without referencing some defined term.  So there is no "offense as defined in"/"offense described in" contrast.  Besides, Congress couldn't have written "an offense defined in 844(i)"; 844(i) is a penalty provision that doesn't define an offense - but does describe one.  

Further, unlike the government, Torres has a contrast-based argument that actually works.  Whereas many of the offenses in the aggravated felony definition are defined generically, Torres's offense was defined by reference to a particular federal statute.  If Congress wanted to make arson sans jurisdictional element an aggravated felony, it need only have generically listed arson.  The government says the generic definition of arson is too confused across the states for that to have worked, but as Torres says, Congress could have provided one Moreover, vague generic offenses didn't trouble Congress when it listed "sexual abuse of a minor" as an aggravated felony.

Torres would seem to have much the better of the arguments at Step One of Chevron, and if he wins at Step One the Court can, if it wants to, duck whether Chevron applies to the Board's interpretation of the aggravated felony definition.  (If he loses at Step One the Court can also duck that question, but I don't see a serious chance of that.)  But if you're getting the sense that this isn't quite a Step One case, you're not alone. Torres's trouble at Step One is that it is, in some instances, natural enough to read "x described in Statute Q" to mean "x less one inapposite aspect of the description in Statute Q."  For example, imagine the following statute:

Section 1: An "exchange" is a health insurance marketplace established by a state that meets the following forty-odd substantive requirements...

Section 2: If a state fails to establish an exchange, HHS shall establish the exchange described in Section 1 and operate it in that state.

Few people, I think, would say that Section 2's command is impossible or even literally nonsensical because an "exchange described in Section 1" must be established by a state.  "Exchange described in Section 1" is naturally read to refer to the substantive requirements of Section 1 only, excluding the identity of the governmental unit that establishes the exchange.

B. Does the Board of Immigration Appeals get Chevron deference?

Complicating what would otherwise be a straightforward Chevron case, the term the Board  has construed in Torres, "aggravated felony," makes key appearances in several immigration crimes.  The most notable of these, as far as real-world prosecutions and real-world impact, is the sentencing provision for illegal reentry - a provision cited in a whopping 16,000 federal appellate decisions, to give readers some sense of the frequency with which illegal reentry prosecutions are brought.*  A deported alien who illegally reenters the country is normally subject to a maximum sentence of two years, but an alien who was deported after (not for) committing an aggravated felony is subject to a maximum sentence of twenty years for reentry.

*Aggravated felonies are also substantive elements of certain immigration crimes.  The most clean example is that while it's a crime to "bring in" to the country all sorts of aliens, it is only a crime to aid or assist an alien to enter the country if he is inadmissible by virtue of an aggravated felony conviction, or by virtue of being a security threat in several respects.  These aid-or-assist prosecutions, however, are rarely brought.

Despite these high criminal-law stakes, the circuits have unanimously given the Board Chevron deference to its interpretation of the aggravated-felony definition, whether in interpreting the immigration or criminal laws where the term appears.  (There have been some murmurings in the courts of appeals that the Board lacks the requisite criminal-law expertise to interpret, even for purely immigration-law purposes, Congress's references to generic offenses and the like, but no circuit has made that its law, and appropriately so, as the Court's "Step Zero" cases on Chevron's applicability have never turned on expertise.)  In the normal course, a circuit defers to the Board on the meaning of a part of the definition in an immigration case, and then, having deferred, applies its deferential precedent in  subsequent criminal cases.

Unanimity in the circuits notwithstanding, the Court has never held whether the Board gets Chevron deference to its interpretation of the aggravated-felony definition; while it's had at least half a dozen cases on various pieces of the definition, those cases have often lacked published precedents of the Board on the issues before the Court.  More problematically for the government, in the last two years, Judge Sutton and Justice Scalia (Chevron's best friend on the Court) have written notable opinions doubting that agencies should receive Chevron deference where they construe language that has both civil and criminal effect.  Torres and a group of Gibson Dunn-represented amici seize on these opinions and argue that the Board can't receive deference to its interpretation of "aggravated felony" because of the definition's substantial criminal law consequences.  Their argument has basically two pieces - first, there can be no Chevron deference to an agency's interpretation of an ambiguous criminal law, because lenity resolves all such ambiguities, and second, only legislatures, not the executive branch, can define crimes.

Assuming that the Court hasn't squarely decided whether agencies get deference to interpretations of statutory language that has criminal-law bite (which I think is basically true, the SG's protestations to the contrary), how should the Court decide this question?  Initially I'm inclined to say that the Court should dodge it in Torres by holding that the Board gets Chevron deference to its interpretations of "aggravated felony" when that term appears in immigration laws, but not when it appears in criminal laws.  Such a move, of course, would have to overcome the obvious objection that a defined term can't mean different things in different places.

Before coming to that objection, a thumbnail sketch of why I think separating civil applications of the aggravated-felony definition from criminal applications is the right move.  Congress didn't delegate authority to the Board to interpret the definition of aggravated felony.  It delegated authority to the Board to interpret deportation statutes that use that defined term.  The Board - a group of political appointees appointed by the Attorney General and her recent predecessors - doesn't have a roving jurisdiction to interpret definitional provisions in the immigration laws.  Rather, it decides, inter alia, appeals from deportation proceedings before immigration judges, and in doing so interprets ambiguous terms in deportation statutes.  One of those ambiguous terms is "aggravated felony," a term that Congress defined with some care, but defined, nevertheless, in a definition shot through with ambiguity and vagueness.  The theory of Chevron is that those ambiguities and vaguenesses represent policy choices that Congress failed to make, and that Congress intended the expert political appointees on the Board, not judges, to make the policy choices left unmade on which criminal aliens should be deported.  On the other hand, Congress obviously did not give the Board authority to decide prosecutions for immigration crimes, and thus under Chevron and Mead the Board has no authority to interpret ambiguous language in statutes defining those crimes or providing sentences for them.

Torres anticipates this move and claims that it "would be impossible" for "aggravated felony" to have two different meanings, since it's a defined term.  But the cases he cites for that proposition, most notably Clark v. Martinez, say that the same provision can't mean different things as applied to different kinds of cases.*   Here, the question is a little different - whether different provisions that use the same term can use that term to mean different things notwithstanding a common definition of the term.  That may sound like the same thing - isn't the definition being given different meanings as applied to different cases? - but it isn't, really.  Here, the claim is that when a defined term is put into different linguistic contexts (not just applied to different facts), those differing contexts act on the ambiguities in the term's definition in different ways and generate different meanings.  

*  As Jonathan Siegel has argued, the Court hasn't even always stuck to this principle.

Whether you buy that distinction or not, the Court does. Just eight years ago in Duke Energy the Court held, with only Justice Thomas in partial dissent, that there is "no effectively irrebuttable presumption that the same defined term in different provisions of the same statute must be interpreted identically.  Context counts."  There, the Court deferred to EPA regulations reading "modification" in one part of the Clean Air Act to mean something other than it does in another part, even though the one provision actually said it was using "modification (as defined in [the other provision])."   

Lest one think otherwise, the Court's nearly unanimous opinion in Duke Energy is no outlier.  In the opinion the Court discusses several recent cases where a defined term was given different meanings as it appeared in different provisions, and Justice Scalia recently cited Duke Energy's seemingly atextual holding with approval in Utility Air by way of holding for the Court that the EPA could and should have given the defined term "air pollutant" different meanings in different parts of the Clean Air Act.  Torres is, if anything, a weaker case for identical interpretation than Duke Energy.  Here, the criminal provisions referring to aggravated felonies don't, of course, talk about aggravated felonies as defined in the deportation statutes.  Whereas in Duke Energy the EPA was the delegated interpreter of both provisions, here the Board is only the delegated interpreter of one of the two provisions at issue.  And in this case, not only is the Board not the delegated interpreter of the criminal provisions, there's a lenity problem and possibly even a separation-of-powers problem (more on which below) with applying the Board's interpretation to those provisions.  

Since the Duke Energy principle is so uncontroversial, applies a fortiori in this case vis-a-vis Duke Energy, and is such an easy way to avoid the tough problem of Chevron deference to interpretations of criminal law, I think there's a pretty decent chance the Court simply cordons off civil interpretations of "aggravated felony" from the term's criminal applications.  There's also, though, a good chance it doesn't.  The government, in arguing the Board should get Chevron deference, has made the curious tactical choice of arguing that the Board's interpretation of the aggravated-felony definition should get Chevron deference in criminal cases, and hasn't argued Duke Energy as a back-up.  There are, in the non-agency context, several cases holding that where a provision of a statute has both civil and criminal applications, lenity requires a globally lenient reading of the statute.  And whereas in the typical Duke Energy case there's some contextual reason that a term in one provision appears to mean something different than it does in another, here the difference isn't one of meaning so much as interpretive authority - arguably making Duke Energy a little inapposite.

Suppose, then, that the Court holds that if the Board gets Chevron deference to its interpretation of the definition of aggravated felony, as that term is used in deportation provisions, courts must then apply that interpretation to criminal statutes' usages of "aggravated felony."  Should the Board still get Chevron deference?  I think the answer's a pretty resounding yes.

The core of the argument against giving the Board Chevron deference, and the core of the Scalia/Sutton argument against giving agencies deference to their interpretations of criminal law, is lenity.  Lenity, the argument goes, resolves all ambiguities in criminal laws, leaving agencies with no gaps to fill.  This is wrong for several reasons.

First, it misunderstands how lenity actually works.  Lenity doesn't kick in whenever a statute is ambiguous in the Chevron Step One sense; it kicks in when, after exhausting every source of statutory meaning, including "history and purpose," a criminal statute remains "grievously" ambiguous.  Now, Chevron is itself a canon of interpretation of sorts; it claims that whenever Congress enacts a statute that is ambiguous (whether or not it's "grievously" so), and gives an agency power to regulate or adjudicate under that statute in a formal enough way, it intends for the statute to bear any reasonable meaning the agency gives it.  Lenity, on the other hand, says that when we can't tell what Congress intended, after every means of answering that question is exhausted, the defendant wins on ambiguities.  Chevron purports to tell us what Congress intended; ergo, it should trump lenity, not the other way around.  (Besides, Chevron is stronger stuff than the inferences from history and purpose that the Court has said numerous times trump lenity.)

Second, lenity, which predates modern administrative agencies by centuries, was developed as a response to judicial interpretation of criminal laws.  Unsurprisingly, its purposes have little bite in the agency context.  Most famously, lenity is about fair warning - the concern that if courts interpret criminal laws to cover conduct they don't clearly cover, defendants will lack adequate notice of their criminal liability.  How does fair notice play in the agency context?  Federal agencies, unlike courts with limited geographical jurisdiction, interpret statutes in regulations or adjudications that apply nationwide.  Really, then, people will usually have better notice of their criminal liability if the meaning of criminal laws is fixed by agency glosses that apply nationwide and interpret a statute with some specificity than they will in the normal situation, where thousands of district judges with local jurisdictions can interpret ambiguous criminal laws as they like, so long as they don't find "grievous" ambiguity.   

Judge Sutton has strained to argue that Chevron deference creates a notice problem because even after Mead agencies can, occasionally, get Chevron deference to some pretty obscure acts of interpretation - policy statements, opinion letters, and the like.  Of course, a patchwork of unpublished district court opinions is equally obscure, if not more so, and I don't see why someone seriously investigating their criminal liability shouldn't be expected to read relevant agency policy statements.  But if obscure policy statements are a serious fair warning concern, that's only a reason to modify Mead so as to not give Chevron deference to agency interpretations of criminal law in policy statements - not a reason to never give Chevron deference to agency interpretations of criminal law.  Torres's amici argue that deferring to agency interpretations of criminal law would create a notice problem because agencies aren't bound by horizontal stare decisis to follow their own prior interpretations.  But neither are district courts.  Stare decisis, horizontal or vertical, doesn't kick in until a court of appeals decides a question.  And even statutory stare decisis doesn't always last forever.  What's more conducive to fair notice - nationally applicable interpretations of criminal law that sometimes change, or numerous locally applicable interpretations of criminal law that, in the aggregate, change a little less?

Lenity is also sometimes said to have a separation-of-powers purpose; lenity, the story goes, ensures that democratically accountable legislatures, not unaccountable courts, define crimes. The advocates of a criminal law carve-out from Chevron say that this purpose of lenity, at least, is served by denying deference to agency interpretations of criminal law.  Agencies, however, aren't courts where accountability is concerned; indeed, the whole premise of Chevron is that they aren't.  What's more, given the realities of lenity doctrine, there's a good argument that Chevron deference would enhance the democratic legitimacy of criminal law, not reduce it.  In a Chevron-free criminal law interpretive regime, completely unaccountable courts are free to make significant policy choices about the content of criminal law so long as they don't find grievous ambiguity.  In my preferred world, it's true that an agency, rather than Congress, could create criminal liability in the rare cases where a statute is so ambiguous that lenity would apply.  (Then again, one could take the position that in those rare cases of grievous ambiguity, lenity should kick in at Step One.)  But in the minerun of cases where there's enough ambiguity for courts to have interpretive leeway, but sufficient material to work with that lenity is off the table, Chevron would move interpretive authority to a more politically accountable, and congressionally controlled, branch of government. 

The opponents of deference to agency interpretations of criminal law also argue that, apart from lenity, there's a freestanding separation-of-powers rule (or really, soft principle) against the executive defining crimes.  Yet they readily concede that Congress may make it a crime to violate a class of regulations that don't yet exist and then authorize an agency to write the regulations.  Judge Sutton suggests that that's different than Chevron, because in that instance Congress has clearly stated an intent to let agencies write criminal law.  He argues a clear-statement rule should govern whether Congress has delegated criminal lawmaking power to agencies.  Justice Scalia, for his part, has vaguely suggested that agency interpretation of existing criminal law is somehow worse than agencies making criminal law from scratch. 

I don't understand Scalia's point at all.  Judge Sutton's argument has, at least, an internal logic to it.  But whatever one's general views on the virtues of courts using clear statement rules to dissuade Congress from doing things it's constitutionally allowed to do (I'm skeptical, obviously), this particular clear statement rule seems a bad idea.  Under the guise of ensuring that only Congress defines crimes unless it clearly states otherwise, it would actually shift criminal-lawmaking authority from mildly politically accountable agency actors to completely unaccountable courts.  As is almost always the case in debates over Chevron's domain, the choice isn't between fidelity to the law Congress made and agencies running amok, but between a political branch making policy and a court doing it.