William Andrew WRIGHT, Petitioner-Appellant, v. Stephen SPAULDING, Warden, Respondent-Appellee.
The circuits have extended habeas law far beyond the limits set by Congress. How far? The simple answer is not far enough to help William Wright. The more complicated answer follows.
First, a quick procedural history. As a felon in possession of a firearm, William Wright would normally receive up to ten years in prison. But Wright had three prior convictions for “serious drug offenses,” so he qualified as an armed career criminal. That meant at least 15 years' imprisonment (and a maximum of life). A Maryland district judge accepted his plea and gave him the minimum sentence. At the time, Wright did not dispute his status as an armed career criminal, nor did he file an appeal.
Instead, Wright challenged his sentence years later, after the Supreme Court handed down Johnson v. United States, ––– U.S. ––––, 135 S. Ct. 2551, 192 L.Ed.2d 569 (2015). That decision held the “residual clause” of the Armed Career Criminal Act to be unconstitutionally vague. Id. at 2563; see 18 U.S.C. § 924(e)(2)(B)(ii). But Wright had a problem: his argument had nothing to do with Johnson or the residual clause (which related to violent felonies, not drug offenses). So the Maryland district court denied his § 2255 motion.
Yet Wright was not done trying to challenge his sentence. He took another shot after the Supreme Court handed down Mathis v. United States, ––– U.S. ––––, 136 S. Ct. 2243, 195 L.Ed.2d 604 (2016). But this time he faced a different problem: he couldn't file a new motion in the sentencing court because the habeas statutes limit “second or successive” motions. See 28 U.S.C. § 2255(h). Instead, Wright filed a habeas corpus petition in the district court where he now happened to be imprisoned: the Northern District of Ohio. The district court dismissed his petition. Wright appealed.
This case is about two things: habeas and holdings. Under the system Congress enacted, Wright's habeas petition would be dead on arrival. Congress prescribed one venue to challenge your sentence after appeal: the sentencing court. And it imposed limits on the number and timing of challenges. But that system left some prisoners without a shot at relief. So courts, this one included, used something known as the “saving clause” to create an escape hatch.
Soon courts found themselves construing not just the law as written but also the law as applied (and misapplied) by courts. Yet interpreting precedents is not always an easy business. Especially when they add to, rather than implement, what Congress has done. To understand what binds us, then, we must first know some basics. About the habeas system: what the statutes say, where they came from, and what about them sent courts looking for a workaround. And about how courts operate: by resolving concrete disputes and announcing the legal rules essential to doing so. Those necessary decisions are the holdings that bind future courts. Not dispensable dicta that sweep more broadly than the issue at hand. To ignore these principles is to risk drifting far from any law enacted by Congress or decided by a court.
Armed with these basics, we discover the law that governs Wright's case. And that law makes clear that Wright's petition must be dismissed.
The general habeas statute empowers federal courts to grant the writ. 28 U.S.C. § 2241(a). Until 1948, any federal prisoner who wanted to challenge his detention relied on this statute. But that created a problem for jurisdictions with many federal prisons, since prisoners had to petition in the district in which they were housed. And that meant certain districts bore the brunt of federal habeas litigation. To solve this problem, Congress enacted § 2255. The new statute directed federal prisoners to challenge their sentences in the district where they were sentenced, not the district of detention. See id. § 2255(a), (c); United States v. Hayman, 342 U.S. 205, 212–19, 72 S.Ct. 263, 96 L.Ed. 232 (1952) (discussing the history); Prost v. Anderson, 636 F.3d 578, 587–88 (10th Cir. 2011) (same). In other words, prisoners would file a motion under § 2255 in their sentencing court, not a traditional habeas petition in the court of their prison district.
As the Supreme Court later explained, the “sole purpose” of § 2255 was to change the venue for challenges to a sentence. Hayman, 342 U.S. at 219, 72 S.Ct. 263. Congress's decision to redirect sentencing back to the sentencing court made perfect sense. If the sentencing judge erred in sentencing the defendant, then he or she should fix it. The best judge to fix a sentence is a judge intimately familiar with the defendant, the case, and the local practices. Not a judge who has never touched the case before.
But § 2255 never replaced § 2241 in its entirety. From the beginning, § 2255 was qualified by a saving clause—in non-legalese, an “unless.” A habeas petition by a federal prisoner is barred “unless ․ the [§ 2255] remedy by motion is inadequate or ineffective to test the legality of his detention.” 28 U.S.C. § 2255(e). In that case, a § 2241 petition might be viable.
The statute does not say when the motion remedy is “inadequate or ineffective[.]” But it is easy to think of examples. By its terms, § 2255 only covers challenges to a sentence. See id. § 2255(a) (authorizing motions “to vacate, set aside, or correct the sentence” “upon the ground that the sentence was imposed in violation of the Constitution or laws of the United States, or that the court was without jurisdiction to impose such sentence, or that the sentence was in excess of the maximum authorized by law, or is otherwise subject to collateral attack”). And an invalid sentence is hardly the only thing a federal prisoner might challenge. Imagine a warden held a prisoner in a manner contrary to or not authorized by the sentence. In that case, § 2255 would be “inadequate or ineffective to test the validity of his detention,” and § 2241 would be the correct cause of action. Id. § 2255(e); see also Samak v. Warden, FCC Coleman-Medium, 766 F.3d 1271, 1284–86 (11th Cir. 2014) (Pryor, J., concurring). Another example would be if the sentencing court no longer exists. See Prost, 636 F.3d at 588; Witham v. United States, 355 F.3d 501, 504–05 (6th Cir. 2004).
That's how courts read the saving clause for most of its history. The rule was simple: § 2255 for attacks on a sentence, § 2241 for other challenges to detention. See, e.g., Wright v. U.S. Bd. of Parole, 557 F.2d 74, 77 (6th Cir. 1977). Then two events led to a judicial expansion of the saving clause.
First, a case. The Supreme Court held that using a firearm required more than mere possession, it required “active employment.” Bailey v. United States, 516 U.S. 137, 144, 116 S.Ct. 501, 133 L.Ed.2d 472 (1995). This holding undid years of more expansive circuit precedent. See id. at 142–43, 116 S.Ct. 501.
Second, a statute. Congress passed the Antiterrorism and Effective Death Penalty Act in 1996. AEDPA tried to curb what courts used to call “abuse of the writ”—the nonstop filing of meritless habeas petitions. See Felker v. Turpin, 518 U.S. 651, 664, 116 S.Ct. 2333, 135 L.Ed.2d 827 (1996). Under the statute, “second or successive” motions by federal prisoners have had to rely on either (1) new and convincing evidence that the prisoner is innocent or (2) a new and previously unavailable rule of constitutional law that the Supreme Court has made retroactive. 28 U.S.C. § 2255(h).
The one-two punch of Bailey and AEDPA raised a problem: what to do about prisoners who (1) were convicted and sentenced under the expansive pre-Bailey definition of using a firearm, (2) filed a § 2255 motion before Bailey, and (3) could not succeed before Bailey because of then-binding circuit precedent. Bailey was not new evidence. Nor was it a new rule of constitutional law. It was just a run-of-the-mill case of statutory interpretation. So Bailey did those prisoners no good—they could not satisfy either of the § 2255(h) gateways for a second or successive motion.
This bothered courts. It seemed unfair to apply § 2255(h) to close the procedural door on prisoners whose Bailey arguments might be winners. And so, one by one, circuits started to devise a workaround: the saving clause. The Third Circuit led the way, holding that § 2255 did not bar a § 2241 petition by someone in the “unusual position” of having had “no earlier opportunity” to bring a Bailey argument. In re Dorsainvil, 119 F.3d 245, 251 (3d Cir. 1997). The Second Circuit followed suit. Triestman v. United States, 124 F.3d 361, 377 (2d Cir. 1997). And the third court to speak, the Seventh Circuit, held that a prisoner had “no reasonable opportunity” to challenge his sentence if binding precedent foreclosed such a challenge. In re Davenport, 147 F.3d 605, 610 (7th Cir. 1998). Since the prisoner had “no reasonable opportunity,” § 2255 was not an adequate remedy. Id. at 611. Enter § 2241 through the saving clause.
By making § 2241 a substitute for procedurally barred § 2255 motions, these cases expanded the saving clause beyond its original function. Still, the expansion was not unbounded. It was a fix for prisoners facing a specific problem: a new case proved their innocence but, practically speaking, they could not obtain relief thanks to binding precedent in the past and procedural barriers going forward. Was that what Congress meant by “inadequate or ineffective” when it wrote the saving clause? Doubtful. But at least the doctrine that crystallized in these earlier cases had a logic—and a limiting principle—of its own. And this logic soon spread to most of the remaining circuits. See, e.g., Abdullah v. Hedrick, 392 F.3d 957, 960–63 (8th Cir. 2004); Ivy v. Pontesso, 328 F.3d 1057, 1059–60 (9th Cir. 2003); Reyes-Requena v. United States, 243 F.3d 893, 904 (5th Cir. 2001); In re Jones, 226 F.3d 328, 333–34 (4th Cir. 2000); Wofford v. Scott, 177 F.3d 1236, 1244 (11th Cir. 1999), overruled by McCarthan v. Dir. of Goodwill Indus.-Suncoast, Inc., 851 F.3d 1076 (11th Cir. 2017) (en banc).
Of course, what binds us here is not the precedent of other circuits but the precedent of this one. And when this circuit joined the bandwagon, it followed the approach then in vogue, letting prisoners use the saving clause to present (1) claims of actual innocence (2) that they could not reasonably present sooner.
But this case is about whether recent decisions have departed from the path laid out by other circuits and our earlier caselaw. See, e.g., Hill v. Masters, 836 F.3d 591 (6th Cir. 2016). And to make that determination, we first must understand what in our prior decisions is binding. Unfortunately, the teachings of precedent are not always as clear as we might wish. Especially in a complicated area like habeas. Most of all when we are dealing with the “rococo jurisprudence” that has built up around the saving clause. Prost, 636 F.3d at 596. So before we take a stroll down precedent lane, we must consider some first principles: What do we look for when we study precedents? What in them are we bound by? And how do we know?
Like most circuits, this circuit follows the rule that the holding of a published panel opinion binds all later panels unless overruled or abrogated en banc or by the Supreme Court. But only holdings are binding, not dicta. Scarber v. Palmer, 808 F.3d 1093, 1096 (6th Cir. 2015); see also Cohens v. Virginia, 19 U.S. (6 Wheat.) 264, 399, 5 L.Ed. 257 (1821). So to know what binds us, we must be able to separate holdings from dicta.
The rule is grounded foremost in constitutional principle. Federal courts are limited to deciding “Cases” and “Controversies.” U.S. Const. art. III, § 2; see also Valley Forge Christian Coll. v. Ams. United for Separation of Church & State, Inc., 454 U.S. 464, 471, 102 S.Ct. 752, 70 L.Ed.2d 700 (1982) (reminding courts that judicial power exists only “to adjudge the legal rights of litigants in actual controversies” (cleaned up)). The federal courts have no power but judicial power. And by its nature, the judicial power does not act on issues in the abstract. See, e.g., Muskrat v. United States, 219 U.S. 346, 361–62, 31 S.Ct. 250, 55 L.Ed. 246 (1911). Rather, it “is capable of acting only when the [issue] is submitted to it by a party who asserts his rights[.]” Osborn v. Bank of the U.S., 22 U.S. (9 Wheat.) 738, 819, 6 L.Ed. 204 (1824). Only then does the issue “become[ ] a case” and “assume such a form that the judicial power is capable of acting on it.” Id.; see also, e.g., Pacificus No. 1 (Alexander Hamilton), in Letters of Pacificus and Helvidius 8 (Washington, Gideon 1845) (“The province of [the judicial] department is to decide litigations in particular cases. It is indeed charged with the interpretation of treaties, but it exercises this function only where contending parties bring before it a specific controversy.”); 1 Baron de Montesquieu, The Spirit of Laws, in The Complete Works 198 (London 1777) (defining “the judiciary power” as that which “punishes criminals, or determines the disputes that arise between individuals”).
The idea of a case or controversy implies a few things.1 One thing it implies is “the existence of present or possible adverse parties, whose contentions are submitted to the court for adjudication.” Muskrat, 219 U.S. at 357, 31 S.Ct. 250 (cleaned up). Another is that the court can do something about the dispute. See Lujan v. Defs. of Wildlife, 504 U.S. 555, 561, 112 S.Ct. 2130, 119 L.Ed.2d 351 (1992); see also Muskrat, 219 U.S. at 362, 31 S.Ct. 250 (explaining that, “[i]n a legal sense, the judgment” of an advisory opinion “could not be executed”). So if the court's answer to a legal question can make no difference to the parties before it, that question is not part of the case. And outside a case, an analysis of an issue can be many things—deliberate or hasty, compelling or unpersuasive, thorough or thinly reasoned—but one thing it cannot be is an exercise of judicial power. Even when it comes packaged in a judicial opinion.
If constitutional principle were not enough, there are also good practical reasons to cabin a precedent's binding force to its holding. As Chief Justice Marshall explained in Cohens, “[t]he question actually before the Court is investigated with care, and considered in its full extent.” 19 U.S. at 399. Collateral issues rarely receive the same treatment. See id. at 399–400. Thus, dictum is less likely to reflect a court's deliberate judgment. See United States v. Burris, 912 F.3d 386, 410 (6th Cir. 2019) (en banc) (Kethledge, J., concurring in the judgment) (“[D]ictum is usually a bad idea, because judges think differently—more carefully, more focused, more likely to think things through—when our words bring real consequences to the parties before us.”). And they may be liable to misinterpretation with snowballing consequences. See Cohens, 19 U.S. at 399–400; United States v. Rabinowitz, 339 U.S. 56, 75, 70 S.Ct. 430, 94 L.Ed. 653 (1950) (Frankfurter, J., dissenting) (describing the “progressive distortion” by which “a hint becomes a suggestion, is loosely turned into dictum and finally elevated to a decision”).
We can now start to distill some working principles. For a court's conclusion about an issue to be part of its holding:
• The decision of the issue must contribute to the judgment: whether and why the court affirms, reverses, vacates, or remands. A legal conclusion that is necessary to the judgment qualifies. So might one sufficient to support the judgment but not strictly necessary in light of an independent and equally sufficient conclusion—that is, when there are two independent reasons for the ruling. See Michael C. Dorf, Dicta and Article III, 142 U. Pa. L. Rev. 1997, 2044 (1994); Michael Abramowicz & Maxwell Stearns, Defining Dicta, 57 Stan. L. Rev. 953, 1027–29 (2005).2 But a conclusion that does nothing to determine the outcome is dictum and has no binding force.
• As a corollary, it must be clear that the court intended to rest the judgment (if necessary) on its conclusion about the issue. Put another way, the court must have actively applied the conclusion to the case in front of it. This rule helps distinguish a case of true independent holdings (e.g., “two genuine issues of material fact independently blocked summary judgment”) from one in which a court decides one issue and merely opines about another (e.g., “because the drivers disagreed about the color of the light, summary judgment was improper; we also note that they disagreed about the time of day, which might matter”).
• Finally, it must be clear that the court considered the issue and consciously reached a conclusion about it. Carroll v. Carroll's Lessee, 57 U.S. 275, 287 [16 How. 275, 14 L.Ed. 936] (1853). This rule honors “the long-held standard that ‘questions which merely lurk in the record, neither brought to the attention of the court nor ruled upon, are not to be considered as having been so decided as to constitute precedents.’ ” Rinard v. Luoma, 440 F.3d 361, 363 (6th Cir. 2006) (alteration adopted) (quoting Nemir v. Mitsubishi Motors Corp., 381 F.3d 540, 559 (6th Cir. 2004)).
Like many legal standards, these principles will sometimes be easier to describe in theory than to apply in practice. But without them, we are sure to fail in either or both of our complementary duties: adhering to precedent when an issue has already been decided and considering an issue with an open mind when it has not.
Now for a look at our own precedents on the saving clause. We joined the bandwagon with Martin v. Perez, 319 F.3d 799 (6th Cir. 2003).3 Like Wright, Martin was a federal prisoner convicted and sentenced by an out-of-circuit district court (the Southern District of Indiana), had filed his initial § 2255 motion in that court (unsuccessfully), and then filed a § 2241 petition in his district of confinement (the Eastern District of Kentucky). See id. at 801–02. Like the prisoners in the post-Bailey cases, he claimed that a new Supreme Court decision proved his innocence by clarifying the elements of his crime of conviction. See id. at 804 (citing Jones v. United States, 529 U.S. 848, 120 S.Ct. 1904, 146 L.Ed.2d 902 (2000)). And like the prisoners in the post-Bailey cases, he would have run into then-binding Seventh Circuit precedent had he raised his innocence argument in an initial § 2255 motion. See United States v. Prevatte, 300 F.3d 792, 795 (7th Cir. 2002) (describing the circuit precedent Jones overruled).
This court agreed that Martin could use the saving clause to pursue his innocence claim. In doing so, it applied the Seventh Circuit's reasoning in Prevatte. Martin, 319 F.3d at 804–05 (citing Prevatte, 300 F.3d at 800). And Prevatte, in turn, applied Davenport's no-reasonable-opportunity test to suggest that saving-clause relief might be available for prisoners in Martin's situation. See Prevatte, 300 F.3d at 798–800. In the same vein, the Martin panel pointed out that “Jones was decided after [Martin] filed his first Section 2255 motion[.]” Martin, 319 F.3d at 805 (citing Prevatte, 300 F.3d at 800). Why? Because the panel thought it mattered that Martin's argument had been blocked by binding precedent until a new Supreme Court case cleared the path.4
So in this circuit, a federal prisoner who has already filed a § 2255 motion and cannot file another one cannot access § 2241 just because a new Supreme Court case hints his conviction or sentence may be defective. Rather, the prisoner must also show that binding adverse precedent (or some greater obstacle) left him with “no reasonable opportunity” to make his argument any earlier, “either when he was convicted and appealed or later when he filed a motion for postconviction relief under section 2255[.]” Davenport, 147 F.3d at 610. Otherwise, § 2255 is simply not inadequate or ineffective to test his claim. And nothing in this court's later precedents gainsays this principle.5
That includes Hill v. Masters, the case Wright and the government emphasize here. In Hill, the panel let a federal prisoner use the saving clause to challenge his pre-Booker sentence under the career-offender sentencing guideline. 836 F.3d at 600; see United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005). And in doing so, it assumed that Hill did not have a previous opportunity to present his claim. Thus, Hill held nothing inconsistent with the reasonable-opportunity standard.
The parties do not focus on the reasonable-opportunity standard when they discuss Hill. Instead, they debate which of two similar, but differently worded, three-part tests in the opinion represents the holding of the case. Compare 836 F.3d at 599–600, with id. at 595. But framing the question that way puts the cart before the horse. Hill's real holding is narrower than either party contends.
The first three-part test, which Wright argues is the holding, comes from the beginning of the opinion's saving-clause discussion:
When seeking to petition under § 2241 based on a misapplied sentence, the petitioner must show (1) a case of statutory interpretation, (2) that is retroactive and could not have been invoked in the initial § 2255 motion, and (3) that the misapplied sentence presents an error sufficiently grave to be deemed a miscarriage of justice or a fundamental defect.
Id. at 595. Under this test, a prisoner must cite a new, retroactive decision and show that it “could not have been invoked in the initial § 2255 motion[.]” Id. To our ears, that sounds like a paraphrase of the no-reasonable-opportunity test that this court inherited from other circuits. Tracing the Hill test to its roots confirms our suspicion. See id. (citing Brown v. Caraway, 719 F.3d 583, 586 (7th Cir. 2013); Williams v. Warden, Fed. Bureau of Prisons, 713 F.3d 1332, 1343 (11th Cir. 2013)); Brown, 719 F.3d at 586 (basing the test on Davenport, 147 F.3d at 611); id. at 587 (holding that “Brown could not have raised his current argument in his first section 2255 motion because it was foreclosed by binding precedent at the time”); Williams, 713 F.3d at 1343 (basing the test on Davenport, 147 F.3d at 611, and its Eleventh Circuit offspring Wofford, 177 F.3d at 1244); id. (“[T]he only sentencing claims that may conceivably be covered by the saving[ ] clause are those based upon a retroactively applicable Supreme Court decision overturning circuit precedent.” (quoting Wofford, 177 F.3d at 1245)); id. (explaining that a “necessary” and “essential” condition of the saving clause “is that the Supreme Court decision must have overturned a circuit precedent that squarely resolved the claim so that the petitioner had no genuine opportunity to raise it at trial, on appeal, or in his first § 2255 motion”).
The Hill panel did not select this test on its own. Rather, the parties “agreed upon” the test that the two other circuits had “set out[.]” 836 F.3d at 595. In other words, all Hill did was apply the law the parties provided.
That was not all the parties agreed on. They also “agree[d]” that Hill “established the first and second conditions[.]” Id. at 595. Unsurprisingly, the panel did not analyze those issues. It just noted, where relevant, that the government “concede[d]” them. Id. at 595–96, 599.
The only point of controversy in Hill was the third element in the parties' test: whether the error in Hill's sentence was a “fundamental defect.” Id. at 595. After considering that issue and the parties' arguments on both sides, the panel said it was. See id. at 596–99. So Hill's § 2241 petition got to pass through the saving clause. Id. at 600.
At no point did the Hill panel consider and consciously decide whether Hill's claim had previously been unavailable. Again, the government conceded that issue and the opinion did not discuss it except to note the concession. Id. at 595. There was no “application of the judicial mind” to that question, so there was no “decision” about it. Carroll, 57 U.S. at 287.
Our hands are not tied in a later case just because, in an earlier one, a party conceded an issue and the panel took that concession at face value. A court's primary duty is to resolve the dispute in front of it. And in our adversarial system, courts chiefly rely on the parties to give them not just the facts but the law and the issues as well. See Carducci v. Regan, 714 F.2d 171, 177 (D.C. Cir. 1983) (Scalia, J.) (“[A]ppellate courts do not sit as self-directed boards of legal inquiry and research, but essentially as arbiters of legal questions presented and argued by the parties before them.”). It would be at least imprudent, and maybe improper, to make binding precedent out of a court's simple acquiescence in the parties' concessions and assumptions. See Cohens, 19 U.S. at 399–400; cf. Massachusetts v. United States, 333 U.S. 611, 639–40, 68 S.Ct. 747, 92 L.Ed. 968 (1948) (Jackson, J., dissenting) (“I see no reason why I should be consciously wrong today because I was unconsciously wrong yesterday.”).
The law of issue preclusion provides a helpful analogy. Under that doctrine, only issues “actually litigated” in a prior case are considered closed in a later one. Restatement (Second) of Judgments § 27 (Am. Law Inst. 1982). And that does not include issues the parties conceded or stipulated to. Id. § 27 cmt. (e); accord United States v. Kasler Elec. Co., 123 F.3d 341, 349 n.13 (6th Cir. 1997). It would be absurd if the rule for stare decisis, which binds our whole circuit, were more sweeping than the rule for issue preclusion, which binds only the parties.6
Ultimately, defining the outer limit of Hill's holding is a question for another day. So too the government's argument that Hill conflicts with yet earlier precedent. Here, it is enough to say that Hill changed absolutely nothing about the reasonable-opportunity requirement that Martin inherited from other circuits. We therefore hold that a federal prisoner cannot bring a claim of actual innocence in a § 2241 petition through the saving clause without showing that he had no prior reasonable opportunity to bring his argument for relief.
Finally, we can turn to this case. Wright bases his habeas petition on Mathis, the 2016 case in which the Supreme Court clarified how to identify ACCA predicate offenses. (The technical name for the method is the “categorical approach.” See Mathis, 136 S. Ct. at 2248.) But as the government points out, Wright's claim never needed Mathis. To be sure, Wright could not cite Mathis, specifically, before it existed. But a claim for habeas relief is more than the talismanic force of a new case name. A new case matters only, if at all, because of the new legal arguments it makes available.7
Labels aside, here is the substance of Wright's claim: properly analyzed under the categorical approach, the Maryland crime he was convicted of in 1989 was not an ACCA-predicate “serious drug offense” because the maximum penalty for that crime was less than ten years in prison. See 18 U.S.C. § 924(e)(2)(A)(ii). Mathis bolsters this argument given that it clarifies the categorical approach. But Mathis did not invent the categorical approach. See Mathis, 136 S. Ct. at 2251 (explaining that Taylor v. United States held that “[a]ll that counts ․ are ‘the elements of the statute of conviction’ ” (quoting 495 U.S. 575, 601, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990))); id. at 2257 (“For more than 25 years, we have repeatedly made clear that application of ACCA involves, and involves only, comparing elements.”). In fact, it did not even break new ground. See id. at 2257 (“Our precedents make this a straightforward case.”); In re Conzelmann, 872 F.3d 375, 376 (6th Cir. 2017) (“The Court's holding in Mathis was dictated by prior precedent (indeed two decades worth).”).
Nor did Wright need Mathis to clear a path through erroneous Fourth Circuit precedent. Wright has identified only one binding Fourth Circuit case (pre- or post-Mathis) that held that his statute of conviction was a serious drug offense. See United States v. Washington, 629 F.3d 403 (4th Cir. 2011). Washington came before Wright's § 2255 motion but well after his sentencing, his opportunity to appeal, and his chance to file a § 2255 motion on any ground within a year of his conviction.8 See 28 U.S.C. § 2255(f)(1).
Wright also points to United States v. Mason, 954 F.2d 219 (4th Cir. 1992). But there, the Fourth Circuit only decided that ACCA does not require a defendant to be convicted of one predicate crime before he commits the next one. Id. at 221. The court did not consider whether the Maryland drug statute was an ACCA predicate in the first place; evidently, that was taken for granted. Thus, under the principles we have already explained, Mason held nothing about that issue. (And, in fact, no court has ever cited Mason as if it did.) In short, Wright has not shown that anything prevented or foreclosed him from making his argument at his sentencing, on direct appeal (had he appealed), or in an initial § 2255 motion.
Thus, Wright had several opportunities to raise his so-called “Mathis claim,” free of any procedural impediments or hostile precedents. That he failed to seize them does not mean that § 2255 was “inadequate or ineffective” to test his sentence. 28 U.S.C. § 2255(e). Having missed those chances, he cannot now use the saving clause to get another bite at the apple.
Whom should we trust to fix a law that may be broken? The people's representatives? Or unelected judges? To the framers and ratifiers of the Constitution, this was an easy question. Our Founders, who knew that “public Virtue is the only Foundation of Republics,” believed in the greatness of the American people. Letter from John Adams to Mercy Warren (Apr. 16, 1775), in 1 The Founders' Constitution 670, 670 (Philip B. Kurland & Ralph Lerner eds., 1987). They trusted the people's representatives to weigh competing interests and make difficult policy choices. That's why they tasked Congress with making the laws and gave life-tenured judges the more modest job of applying them.
But sometimes we exceed that modest role. Take the saving clause. Section 2255 is not “inadequate or ineffective to test the legality of [a prisoner's] detention” just because binding circuit precedent is against him at the time of his conviction or his first motion. 28 U.S.C. § 2255(e). Applying AEDPA in such cases may seem unfair. Yet in enacting AEDPA, Congress weighed error correction against finality and made some difficult policy judgments. See Prost v. Anderson, 636 F.3d 578, 582–83 (10th Cir. 2011) (noting the real-world benefits of finality “to victims, their families, to future potential victims, to the government, and to the courts”). And for us to second-guess those judgments is to usurp Congress's role.
Not all usurpation is willful. Sometimes it happens almost by accident. For instance, this court followed other circuits without once stopping to think whether it was taking a wrong turn. We started by assuming that the out-of-circuit cases might be right. See United States v. Peterman, 249 F.3d 458, 462 (6th Cir. 2001); Charles v. Chandler, 180 F.3d 753, 757 (6th Cir. 1999). Soon, we assumed they were right. See Martin v. Perez, 319 F.3d 799, 803–04 (6th Cir. 2003). But at no point did we stop to study the text of the statute itself. Justice Frankfurter's lament comes to mind: “a hint becomes a suggestion, is loosely turned into dictum and finally elevated to a decision.” United States v. Rabinowitz, 339 U.S. 56, 75, 70 S.Ct. 430, 94 L.Ed. 653 (1950) (Frankfurter, J., dissenting).
Within the last decade, one circuit unencumbered by precedent took a fresh look at the saving clause and declined to follow the herd. See Prost, 636 F.3d 578. A second broke ranks after years of struggling to apply its old rule coherently. See McCarthan v. Dir. of Goodwill Indus.-Suncoast, Inc., 851 F.3d 1076 (11th Cir. 2017) (en banc); see also id. at 1100–01 (Carnes, C.J., concurring). Those opinions and a few others explain in detail why restarting the collateral-review process every time the Supreme Court overturns circuit precedent is atextual, illogical, and unworkable. See United States v. Surratt, 797 F.3d 240, 251–55, 257–263 (4th Cir. 2015), vacated on grant of reh'g en banc, appeal dismissed as moot, 855 F.3d 218 (4th Cir. 2017); Samak v. Warden, FCC Coleman-Medium, 766 F.3d 1271, 1279–93 (11th Cir. 2014) (Pryor, J., concurring); Brown v. Caraway, 719 F.3d 583, 597–600 (7th Cir. 2013) (Easterbrook, C.J., statement concerning the circulation under Circuit Rule 40(e)). There is plenty to read for those who want the full argument. No need to rehash everything here. Instead, I will just expand on two points. First, a word about practicality. Second, a word about precedent.
Practicality. Judicial solutions often beget judicial problems. The way courts have kludged the saving clause is a perfect example. Using § 2241 to replace procedurally barred § 2255 motions channels federal prisoners' postconviction challenges back into the traditional habeas system. And that means resurrecting the very problems § 2255 was supposed to put to rest. See John F. Manning, What Divides Textualists from Purposivists?, 106 Colum. L. Rev. 70, 84 (2006) (“[T]extualists recognize that the relevant context for a statutory text includes the mischiefs the authors were addressing.”). Concentrating “an inordinate number of habeas corpus actions” in districts with large prison populations? United States v. Hayman, 342 U.S. 205, 213–14, 72 S.Ct. 263, 96 L.Ed. 232 (1952). Check.1 Asking district courts to review each other's proceedings—often without access to the witnesses, the sources of evidence, or other local information that may be critical? See id. at 213, 72 S.Ct. 263. Check again.
All this was bad enough to begin with. But it became even worse after Hill v. Masters, which expanded the saving clause to cover not just invalid sentences but miscalculated ones too. See 836 F.3d 591 (6th Cir. 2016). In the former case, the remedy is usually simple: an order for the prisoner's release. But in the latter, only a resentencing will do. Query how the Great Writ of habeas corpus—an order served on the respondent personally to determine the validity of his custody over the petitioner—can be transmuted into a resentencing vehicle at all. See Braden v. 30th Judicial Circuit Court of Ky., 410 U.S. 484, 494–95, 93 S.Ct. 1123, 35 L.Ed.2d 443 (1973). But even bracketing that conundrum, how exactly is a district judge in Kentucky supposed to know the right sentence for a South Carolinian who conspired to traffic drugs in Charleston? District judges must weigh several sentencing factors in the scales of justice. 18 U.S.C. § 3553(a). And many of those factors can only be weighed meaningfully if the judge knows something about the defendant, the case, and the local community. No wonder, then, that Congress directed sentencing challenges to the sentencing court.
It should be no surprise when judicial creativity begets chaos. Judges are not cut out for legislative craftsmanship. The Founders understood this. In fact, they rejected the Council of Revision, a proposal to give the judiciary a veto over the laws, for this very reason. Simply put, they did not trust judges to make policy. See 1 The Records of the Federal Convention of 1787 97–98 (Max Farrand ed., 1911) (Elbridge Gerry, June 4) (“It was quite foreign from the nature of [the judicial] office to make them judges of the policy of public measures.”); 2 id. at 73 (Nathaniel Gorham, July 21) (“As Judges they are not to be presumed to possess any peculiar knowledge of the mere policy of public measures.”); id. at 76 (Luther Martin, July 21) (“A knowledge of Mankind, and of Legislative affairs cannot be presumed to belong in a higher [ ] degree to the Judges than to the Legislature.”); see also John F. Manning, Textualism and the Equity of the Statute, 101 Colum. L. Rev. 1, 61 (2001) (noting that the Founders' “sharp separation of legislative and judicial powers was designed, in large measure, to limit judicial discretion”). But if courts can't resist tinkering with the law, the least they can do is to spare a thought for the consequences. Not just for the present case, but also for future cases and for how later courts will have to implement the tinkering as a precedent. Which brings me to my next point.
Precedent. Precedent matters. This case shows that in more ways than one. Indeed, it's what launched this whole saving-clause misadventure in the first place. Think back to In re Davenport. 147 F.3d 605 (7th Cir. 1998). The premise of that case was that court proceedings are “inadequate or ineffective to test the legality of” a conviction or sentence if binding-but-wrong precedent dictates the outcome—at least, at the district-court and panel levels. 28 U.S.C. § 2255(e). In Davenport's words, binding precedent leaves defendants without “a reasonable opportunity to obtain a reliable judicial determination” of their case, without “any opportunity for judicial rectification” of mistakes. 147 F.3d at 609, 611.
There's some appeal to this reasoning. No one wants prisoners to stay in prison longer than the law says they should. And there does seem to be something unfair about a circuit court—the last court to which a defendant can appeal as of right—closing its ears to what should be a winning argument in deference to a precedent that is just plain wrong.2
But here's the problem: In our system, litigants are bound by precedent all the time. Even precedent that we later recognize was incorrect. In this circuit, when our precedent cuts against a party, we expect that party to (1) distinguish it, (2) persuade us to overrule it en banc, or (3) persuade the Supreme Court to correct our error. In the meantime, we go on applying it. And an overruling down the line does not blot out final judgments based on that precedent back when it was binding.3 Does this mean that our entire justice system (in which, again, precedent is ubiquitous) is “inadequate or ineffective” to test the rights of parties? Surely not. And if it somehow did, the problem would not at all be specific to criminal law.
So whatever pull Davenport's reasoning may have, it is not an argument for how courts have construed § 2255(e). The saving clause originally stood for a simple idea: while § 2255 governs attacks on the sentence, § 2241 still governs attacks on its execution. When Congress added AEDPA's second-and-successive limitations without touching the text of § 2255(e), did it implicitly amend the saving clause to now stand for the unrelated (and radical) idea that binding precedent alone is enough to strip a full round of judicial process—trial, appeal, and one free collateral attack—of its basic adequacy? Of course not. See Antonin Scalia & Bryan A. Garner, Reading Law: The Interpretation of Legal Texts § 39, at 255 (2012) (“This would be repealer by the weakest of implications; and repeals by implication are disfavored[.]”); see also id. § 40, at 256 (discussing objective statutory history's relevance to meaning). Talk about elephants in mouseholes.
Section 2255 is not that complicated. A federal prisoner can attack his sentence with a motion to correct or vacate it under § 2255—subject to the limitations Congress has imposed, including the restrictions on second or successive motions. Or he can challenge the manner of his detention with a habeas petition under § 2241—just as the saving clause of § 2255(e) allows. But we have jerry-rigged the saving clause to let § 2255 motions masquerade as § 2241 petitions, evading Congress's chosen limits on collateral review. This even though “judgments about the proper scope of the writ are normally for Congress to make.” Felker v. Turpin, 518 U.S. 651, 664, 116 S.Ct. 2333, 135 L.Ed.2d 827 (1996) (cleaned up). And we are bound by our mistakes in this area until we fix them en banc.
Or until the Supreme Court intervenes. If this circuit and others fail to course-correct on our own, then the Court should step in. And I would respectfully submit that sooner may be better than later. The circuits are already split. The rift is unlikely to close on its own. What's more, so long as it lasts, the vagaries of the prison lottery will dictate how much postconviction review a prisoner gets. A federal inmate in Tennessee can bring claims that would be thrown out were he assigned to neighboring Alabama. Like cases are not treated alike.
Maybe worst of all, Congress can hardly deliberate about whether the law should change if no one has a clear idea of what the law is now. Which is one more reason the judiciary should stay in its lane. When we apply the law as written, we show Congress the results of its work and allow it to decide whether they are satisfactory. By doing our job—and nothing more—we most help the people's representatives to do theirs. Exactly as the Framers intended.
1. We asked the parties to file supplemental briefs on the original meaning of Article III's case-or-controversy requirement, specifically whether the corpus of Founding-era American English helped illuminate that meaning. A team of corpus linguistics researchers submitted two amicus briefs as well. We are grateful to both the parties and the amici for their hard work. Here, we agree with the parties that corpus linguistics turned out not to be the most helpful tool in the toolkit. See Wilson v. Safelite Grp., Inc., 930 F.3d 429, 445–46 (6th Cir. 2019) (Stranch, J., concurring) (voicing some practical reservations about corpus linguistics); id. at 440 (Thapar, J., concurring in part and concurring in the judgment) (agreeing that “corpus linguistics is one tool ․ but not the whole toolbox”).
2. The same may go for certain conclusions that, while not affecting the judgment by themselves, are necessary to structure the court's path to those conclusions that are decisive for the judgment. But this is a hard question. Compare, e.g., Dorf, supra, at 2046 (arguing that a court's “considered opinion” that an error occurred is not dictum, even if the court holds that it was harmless, in part because “until the court passes on the substantive question, it will not know exactly what the error is that it must test for harmlessness”), with Pierre N. Leval, Judging Under the Constitution: Dicta About Dicta, 81 N.Y.U. L. Rev. 1249, 1268 (2006) (condemning as dictum the “error” part of “error, but harmless error”), and with Abramowicz & Stearns, supra, at 1065 (agreeing with Dorf, but pinpointing “exactly why the harmless-error question is difficult, namely that one might dispute just how ordered the test is”). And luckily, it is one we need not resolve here.
3. In two earlier cases, this court denied saving-clause access while leaving it an open question whether it would follow the other circuits' approach. See United States v. Peterman, 249 F.3d 458, 460, 462 (6th Cir. 2001); Charles v. Chandler, 180 F.3d 753, 757 (6th Cir. 1999). And in a still earlier case, it opined that any procedural barrier erected by AEDPA would make § 2255 inadequate or ineffective. See In re Hanserd, 123 F.3d 922, 929–30 (6th Cir. 1997). But that statement was clearly dictum because the court resolved the case on the simple ground that AEDPA did not apply. See id. at 933–34.
4. Martin also “affirmed the authority” of Early v. Lamanna, an unpublished case that “more fully” laid out the ideas underpinning Martin. 319 F.3d at 803 n.1 (citing No. 98-3892, 1999 WL 435156 (6th Cir. June 17, 1999)). And Early was even clearer that to file a § 2241 petition through the saving clause, a prisoner must show “circumstances that prevented him from asserting his claims at an earlier opportunity,” such as “an intervening change in the law” of the kind Davenport, Triestman, and Dorsainvil all involved. 1999 WL 435156, at *5 (citing Davenport, 147 F.3d at 611–12; Triestman, 124 F.3d at 377–80; Dorsainvil, 119 F.3d at 251–52); see also id. at *4, 6.
5. True, some cases have glossed over this requirement. But not all. See Wooten v. Cauley, 677 F.3d 303, 307 (6th Cir. 2012) (noting in dicta that the petitioner “could not have raised [a case narrowing an element of the crime] in his prior proceedings”). And the ones that do gloss over it hold nothing to the contrary; in those cases, the issue was not considered nor would it have changed any outcomes. See Harrington v. Ormond, 900 F.3d 246, 249 (6th Cir. 2018) (approving § 2241 petition by prisoner who was sentenced in Iowa and relied on a Supreme Court case overruling adverse Eighth Circuit precedent); Paulino v. United States, 352 F.3d 1056, 1061 (6th Cir. 2003) (dismissing § 2241 petition not based on actual innocence); Bannerman v. Snyder, 325 F.3d 722, 724 (6th Cir. 2003) (same).
6. To be clear, we are not taking a hard line on the complicated issue of whether, or to what extent, or in what circumstances, parties can “stipulate the law.” See generally Gary Lawson, Stipulating the Law, 109 Mich. L. Rev. 1191 (2011). Courts clearly have the power to determine the governing law independent of the parties' representations. See U.S. Nat'l Bank of Or. v. Indep. Ins. Agents of Am., Inc., 508 U.S. 439, 446–48, 113 S.Ct. 2173, 124 L.Ed.2d 402 (1993). But, just as clearly, they do not always exercise that power. See, e.g., Free Enter. Fund v. Pub. Co. Accounting Oversight Bd., 561 U.S. 477, 483, 130 S.Ct. 3138, 177 L.Ed.2d 706 (2010) (“The parties do not ask us to reexamine any of these precedents, and we do not do so.”); id. at 487, 130 S.Ct. 3138 (“The parties agree that [SEC] Commissioners [are not removable at will], and we decide the case with that understanding.” (citations omitted)). For example, in Free Enterprise Fund, when the Supreme Court assumed (based on the parties' agreement) that SEC commissioners are only removable for cause and “decide[d] the case with that understanding,” it surely did not create binding precedent to that effect. Id. at 487, 130 S.Ct. 3138. The same principle applies to this court's decisions.
7. Sometimes courts have phrased the saving-clause test in terms of whether a Supreme Court case announced a “new rule.” See, e.g., Chazen v. Marske, 938 F.3d 851, 861–62, 2019 WL 4254295, at *8–9 (7th Cir. 2019) (noting confusing variations in how the Seventh Circuit's precedents have stated the test); id. at 865–66, 2019 WL 4254295, at *12–13 (Barrett, J., concurring) (same). Yet this phrasing risks confusing the test for saving-clause access with Teague v. Lane's “new rule” test for retroactivity, a separate habeas-related doctrine. See 489 U.S. 288, 301, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989). This “new rule” concept is ubiquitous in habeas law—both because Teague is an important habeas doctrine and because some AEDPA provisions incorporate it by reference. See, e.g., 28 U.S.C. § 2255(h)(2). So courts sometimes fall back on this familiar “new rule” language in the context of saving-clause cases. But luckily our circuit has avoided this mistake. The post-Bailey-and-AEDPA saving-clause caselaw is not based on Teague. Nor does it interpret any provisions of AEDPA. (On the contrary, it exists to circumvent AEDPA.) Rather than mirroring Teague's “new rule” test, the central idea of the post-Bailey-and-AEDPA saving-clause cases is “whether an intervening case of statutory interpretation opens the door to a previously foreclosed claim.” Chazen, 938 F.3d at 862, 2019 WL 4254295, at *9.
8. What's more, Washington complies with Mathis and all other Supreme Court caselaw. The Fourth Circuit in Washington used the “modified categorical approach,” which involves looking at certain state-court records to pin down a conviction under a “divisible” statute—one that defines multiple crimes with distinct sets of elements. See 629 F.3d at 408, 413–14. And the Maryland “possession with intent to distribute” statute was divisible. It laid out sentencing ranges with maximum prison terms that varied depending on the schedule of the controlled substance. See id. at 408. Which means that the identity of the substance (or at least that it belonged to a schedule matching the right sentencing range) was an element of the offense. See Apprendi v. New Jersey, 530 U.S. 466, 490, 120 S.Ct. 2348, 147 L.Ed.2d 435 (2000). Which means, in turn, that the statute was divisible and the Fourth Circuit correctly applied the modified categorical approach. See Mathis, 136 S. Ct. at 2248–49.
1. Seven federal districts house more than 20% of all federal inmates: the Eastern District of Kentucky, the Southern District of Mississippi, the Eastern District of North Carolina, the Western District of Oklahoma, the Middle District of Pennsylvania, and the Northern and Southern Districts of West Virginia. See Federal Bureau of Prisons, Population Statistics, https://www.bop.gov/about/statistics/population_statistics.jsp (last visited Sept. 18, 2019); Federal Bureau of Prisons, Map of our Locations, https://www.bop.gov/locations/map.jsp (last visited Sept. 18, 2019). But those seven courts account for only about 5% of the nation's seats for active district judges. See Administrative Office of the United States Courts, Additional Authorized Judgeships–Since 1960, https://www.uscourts.gov/sites/default/files/districtauth.pdf (last visited Sept. 18, 2019).
2. One more reason why it is so important for courts to understand what prior decisions actually held. See Maj. Op., supra, at 700–03.
3. In exceptional circumstances, courts may cut a party some slack based on intervening changes in caselaw. For example, habeas petitioners can sometimes revive forfeited arguments too novel to be reasonably available at the time of their forfeiture. Thus, novelty may satisfy the “cause” half of the cause-and-prejudice test to cure a procedural default. See Reed v. Ross, 468 U.S. 1, 16, 104 S.Ct. 2901, 82 L.Ed.2d 1 (1984). But even Reed's generous exception is not as generous as the majority approach to the saving clause: under Reed, futility stemming from adverse circuit precedent is not “cause.” See Bousley v. United States, 523 U.S. 614, 623, 118 S.Ct. 1604, 140 L.Ed.2d 828 (1998); Smith v. Murray, 477 U.S. 527, 534–37, 106 S.Ct. 2661, 91 L.Ed.2d 434 (1986). In the Supreme Court's words, “futility cannot constitute cause if it means simply that a claim was unacceptable to that particular court at that particular time.” Bousley, 523 U.S. at 623, 118 S.Ct. 1604 (cleaned up). The Court adopted this view precisely because it recognized that courts can, and do, rethink their own precedents. See Smith, 477 U.S. at 534–35, 106 S.Ct. 2661; Engle v. Isaac, 456 U.S. 107, 130, 102 S.Ct. 1558, 71 L.Ed.2d 783 (1982). Courts seem to have forgotten this basic point in the post-AEDPA world.
THAPAR, Circuit Judge.
THAPAR, J., delivered the opinion of the court in which SILER, J., and HOOD, D.J., joined. THAPAR, J. (pp. 706–10), also delivered a separate concurring opinion.