concurring.
Anyone who violates the Real Estate Settlement Procedures Act’s ban on referral fees commits a crime. See 12 U.S.C. § 2607(d)(1). The rule of lenity tells all interpreters to resolve uncertainties in laws with criminal applications in favor of the defendant. But the Department of Housing and Urban Development has resolved an ambiguity in the law against the defendant, and the government insists that we must defer to this understanding. The doctrine of Chevron deference, the government explains, leaves us no choice. This theory would allow one administration to criminalize conduct within the scope of the ambiguity, the next administration to decriminalize it, and the third to recriminalize it, all without any direction from Congress. I am skeptical.
The court does not go into detail in exploring how the rule of lenity interacts with Chevron because the issue does not drive the outcome of this case. But because this question will return sooner or later, I write to offer some thoughts on how to address it when it does.
The rule of lenity tells courts to interpret ambiguous criminal laws in favor of criminal defendants. United States v. Wiltberger, 5 Wheat. 76, 95, 5 L.Ed. 37 (1820). This principle rests on concerns about notice (the state ought to provide fair warning of what violates the criminal laws) and separation of powers (Congress, not agencies or courts, defines crimes). United States v. Bass, 404 U.S. 336, 348, 92 S.Ct. 515, 30 L.Ed.2d 488 (1971). The Chevron doctrine tells courts to defer to an administrative agency’s reasonable interpretation of an ambiguous statute. Chevron U.S.A., Inc. v. Natural Resources Defense Council, 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). This principle rests on the presumption that, when Congress leaves a statutory gap, it means for the agency rather than the court to fill it. Id. at 843-44,104 S.Ct. 2778.
*730The two rules normally operate comfortably in their own spheres. The rule of lenity has no role to play in interpreting humdrum regulatory statutes, which contemplate civil rather than criminal enforcement. And Chevron has no role to play in interpreting ordinary criminal statutes, which are “not administered by any agency but by the courts.” Crandon v. United States, 494 U.S. 152, 177, 110 S.Ct. 997, 108 L.Ed.2d 132 (1990) (Scalia, J., concurring in the judgment); see Gonzales v. Oregon, 546 U.S. 243, 264, 126 S.Ct. 904, 163 L.Ed.2d 748 (2006).
What happens with a hybrid statute? Today’s Act imposes civil and criminal penalties for violating the provision at issue. See 12 U.S.C. § 2607(d). And it empowers an executive agency to administer the provision by making rules and holding hearings. See id. § 2617. As between the rule of lenity and the agency’s interpretation, which one resolves statutory doubt?
One possibility is to apply the rule of lenity in criminal prosecutions and to defer to the agency’s position in civil actions. But a statute is not a chameleon. Its meaning does not change from case to case. A single law should have one meaning, and the “lowest common denominator, as it were, must govern” all of its applications. Clark v. Martinez, 543 U.S. 371, 380, 125 S.Ct. 716, 160 L.Ed.2d 734 (2005).
United States v. Thompson/Center Arms Co. illustrates the point. The Court had to interpret a law that included a civil tax penalty and a criminal penalty. Even though Thompson/Center Arms was a tax case, the Court applied the rule of lenity. 504 U.S. 505, 518 n. 10, 112 S.Ct. 2102, 119 L.Ed.2d 308 (1992) (plurality opinion); id. at 519, 112 S.Ct. 2102 (Scalia, J., concurring in the judgment). “The rule of lenity,” the lead opinion explained, “is a rule of statutory construction^] ... not a rule of administration calling for courts to refrain in criminal cases from applying statutory language that would have been held to apply if challenged in civil litigation.” Id. at 518 n. 10, 112 S.Ct. 2102 (plurality opinion). Recent cases reaffirm the point. See, e.g., Maracich v. Spears, — U.S. -, 133 S.Ct. 2191, 2209, 186 L.Ed.2d 275 (2013); Kasten v. Saint-Gobain Performance Plastics Corp., — U.S. --, 131 S.Ct. 1325, 1336, 179 L.Ed.2d 379 (2011); Leocal v. Ashcroft, 543 U.S. 1, 11 n. 8, 125 S.Ct. 377, 160 L.Ed.2d 271 (2004); Scheidler v. Nat’l Org. for Women, 537 U.S. 393, 408-09, 123 S.Ct. 1057, 154 L.Ed.2d 991 (2003).
Case law thus makes clear that either the rule of lenity prevails across the board or the agency’s interpretation does. But which one? The better approach, it seems to me, is that a court should not defer to an agency’s anti-defendant interpretation of a law backed by criminal penalties.
First, the rule of lenity forbids deference to the executive branch’s interpretation of a crime-creating law. If an ordinary criminal law contains an uncertainty, every court would agree that it must resolve the uncertainty in the defendant’s favor. No judge would think of deferring to the Department of Justice. Allowing prosecutors to fill gaps in criminal laws would “turn the normal construction of criminal statutes upside down, replacing the doctrine of lenity with a doctrine of severity.” Crandon, 494 U.S. at 178, 110 S.Ct. 997 (Scalia, J., concurring in the judgment).
If the rule of lenity forecloses deference to the Justice Department’s interpretation of a crime-creating law in Title 18, does it not follow that it forecloses deference to the Housing Department’s interpretation of a crime-creating law in Title 12? Or the immigration authorities’ interpretation of a crime-creating law in Title 8? Or the IRS’s *731interpretation of a crime-creating law in Title 26? No principled distinction separates these settings. Allowing housing inspectors and immigration officers and tax collectors to fill gaps in hybrid criminal laws, no less than allowing prosecutors to fill them in pure criminal laws, offends the rule of lenity.
Second, looking at the question within the framework of Chevron leads to the same answer. An agency’s interpretation of a statute does not prevail whenever the face of the statute contains an ambiguity. Deference comes into play only if a statutory ambiguity lingers after deployment of all pertinent interpretive principles. If you believe that Chevron has two steps, you would say that the relevant interpretive rule — the rule of lenity — operates during step one. Once the rule resolves an uncertainty at this step, “there [remains], for Chevron purposes, no ambiguity ... for an agency to resolve.” INS v. St. Cyr, 533 U.S. 289, 320 n. 45, 121 S.Ct. 2271, 150 L.Ed.2d 347 (2001). If you believe that Chevron has only one step, you would say that Chevron requires courts “to accept only those agency interpretations that are reasonable in light of the principles of construction courts normally employ.” EEOC v. Arabian American Oil Co., 499 U.S. 244, 260, 111 S.Ct. 1227, 113 L.Ed.2d 274 (1991) (Scalia, J., concurring in part and concurring in the judgment). If an interpretive principle resolves a statutory doubt in one direction, an agency may not reasonably resolve it in the opposite direction. Id. But the broader point, the critical one, transcends debates about the mechanics of Chevron: Rules of interpretation bind all interpreters, administrative agencies included. That means an agency, no less than a court, must interpret a doubtful criminal statute in favor of the defendant.
Precedents in related areas confirm this conclusion. All manner of presumptions, substantive canons and clear-statement rules take precedence over conflicting agency views.. See Wyeth v. Levine, 555 U.S. 555, 576, 129 S.Ct. 1187, 173 L.Ed.2d 51 (2009) (presumption against preemption); St. Cyr, 533 U.S. at 320, 121 S.Ct. 2271 (presumption against retroactivity); id. (interpretation of doubtful deportation statutes in favor of immigrants); Alexander v. Sandoval, 532 U.S. 275, 288-91, 121 S.Ct. 1511, 149 L.Ed.2d 517 (2001) (presumption against implied causes of action); SWANCC v. U.S. Army Corps of Eng’rs, 531 U.S. 159, 173, 121 S.Ct. 675, 148 L.Ed.2d 576 (2001) (federalism canon); Edward J. DeBartolo Corp. v. Florida Gulf Coast Bldg. & Constr. Trades Council, 485 U.S. 568, 575, 108 S.Ct. 1392, 99 L.Ed.2d 645 (1988) (avoidance of constitutional doubt). Why treat the rule of lenity, the most venerable and venerated of interpretive principles, differently?
Third, the policies that drive lenity and Chevron show how to harmonize the two principles. Start with lenity. Making something a crime is serious business. It visits the moral condemnation of the community upon the citizen who engages in the forbidden conduct, and it allows the government to take away his liberty and property. The rule of lenity carries into effect the principle that only the legislature, the most democratic and accountable branch of government, should decide what conduct triggers these consequences. Bass, 404 U.S. at 348, 92 S.Ct. 515. By giving unelected commissioners and directors and administrators carte blanche to decide when an ambiguous statute justifies sending people to prison, the government’s theory diminishes this ideal.
The rule of lenity also compels the state to give the citizen fair warning, ideally on the face of the statute, of what the criminal
*732law forbids. McBoyle v. United States, 283 U.S. 25, 27, 51 S.Ct. 340, 75 L.Ed. 816 (1931). There are no crimes by implication just as no one is killed by implication. Yet if agencies are free to ignore the rule of lenity, the state could make an act a crime in a remote statement issued by an administrative agency. The agency’s pronouncement need not even come in a notice-and-comment rule. All kinds of administrative documents, ranging from manuals to opinion letters, sometimes receive Chevron deference. See, e.g., Barnhart v. Walton, 535 U.S. 212, 221-22, 122 S.Ct. 1265, 152 L.Ed.2d 330 (2002). Nor is this a figment. In this case, the government has tried to expand a federal criminal law through a policy statement, a theory that runs headlong into “the instinctive distastes against men languishing in prison unless the lawmaker has clearly said they should.” Henry Friendly, “Mr. Justice Frankfurter and the Reading of Statutes,” in Benchmarks 196, 209 (1967).
So much for the purpose of lenity; what of the purpose of Chevron? There may be as many accounts of Chevron as there are professors of administrative law. But what matters most, Chevron’s account of itself, shows that Chevron accommodates rather than trumps the lenity principle. Filling a statutory gap, the Supreme Court explained, requires making a policy choice. 467 U.S. at 864-65, 104 S.Ct. 2778. But courts should avoid making policy choices, as they enjoy neither expertise in the relevant area nor a democratically accountable pedigree. Id. at 865-66, 104 S.Ct. 2778. Forced to a choice between the two, the Court concluded that administrators are better equipped than judges to fill the gaps. Id. at 866,104 S.Ct. 2778.
This account of Chevron says nothing about the present case. • When a court applies the rule of lenity, it does not snatch a policy decision from the political branches. It instead insists that the choice to make the conduct criminal be made by the first political branch rather than the second. Put another way, Chevron describes how judges and administrators divide power. But power to define crimes is not theirs to divide. The accommodation then becomes straightforward: Allowing agencies to fill gaps in criminal statutes would impair the rule of lenity’s purposes, and interpreting these statutes leniently would respect Chevron’s aims.
Fourth, uninvited oddities arise if courts but not agencies must adhere to the rule of lenity. United States v. Mead Corp., 533 U.S. 218, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001), and its follow-on cases hold that an agency interpretation’s eligibility for Chevron deference depends on the procedure that preceded the interpretation’s adoption as well as on factors like “the interstitial nature of the legal question, the related expertise of the Agency, the importance of the question to the administration of the statute, the complexity of that administration, and the careful consideration the Agency has given the question over a long period of time.” Walton, 535 U.S. at 222, 122 S.Ct. 1265. Where the governing statute creates only civil liability, a multi-factor test may be the best one can hope for. See Mead, 533 U.S. at 236-37, 121 S.Ct. 2164. But it is a bit strange to say that, if Welles-Bowen wants to know whether it commits a crime by falling afoul of a policy statement, it must first endure the “open-ended rough-and-tumble of factors.” Medellin v. Texas, 552 U.S. 491, 514, 128 S.Ct. 1346, 170 L.Ed.2d 190 (2008).
Auer v. Robbins, 519 U.S. 452, 117 S.Ct. 905, 137 L.Ed.2d 79 (1997), adds another complication. It says that, when a regulation interpreting an ambiguous statute itself contains an ambiguity, the agency’s interpretation of the regulation receives essentially complete deference. See id. at *733461, 117 S.Ct. 905; Bowles v. Seminole Rock & Sand Co., 325 U.S. 410, 414, 65 S.Ct. 1215, 89 L.Ed. 1700 (1945). Unless the rule of lenity applies to agencies, Auer would give each agency two ways of construing criminal laws against the defendant — by resolving ambiguities in the criminal statute and by resolving ambiguities in any regulation. What’s more, the range of documents eligible for deference under Auer is broader than under Chevron. Even an interpretation contained in a brief may receive deference. Auer, 519 U.S. at 462, 117 S.Ct. 905. One head-turning upshot of permitting Chevron to silence the rule of lenity is this: Any government lawyer with a laptop could create a new federal crime by adding a footnote to a friend-of-the-eourt brief. That is not likely.
The retroactivity of Chevron deference adds another paradox. An agency’s authoritative interpretation of a statute attracts deference even in cases about transactions that occurred before the issuance of the interpretation. Smiley v. Citibank (South Dakota), N.A., 517 U.S. 735, 744 n. 3, 116 S.Ct. 1730, 135 L.Ed.2d 25 (1996). But how would this rule work in a criminal setting given the Ex Post Facto Clause? See U.S. Const, art. I, § 9, cl. 3. So long as the one-statute, one-interpretation rule stands, a court cannot dignify the one principle without slighting the other.
The government, both in this case and in similar cases before other courts, offers several lines of argument in response to this approach. None is convincing.
The government points out that several cases show that Congress’s authority to define crimes is not exclusive. Although the Constitution as a general matter vests power to define crimes in Congress alone, the modern nondelegation doctrine, it is true, occasionally allows Congress to transfer some responsibility for defining crimes to the executive branch. Hence United States v. Grimaud, 220 U.S. 506, 31 S.Ct. 480, 55 L.Ed. 563 (1911), held that Congress could make it a crime to violate regulations issued by the Secretary of Agriculture. Touby v. United States, 500 U.S. 160, 111 S.Ct. 1752, 114 L.Ed.2d 219 (1991), held that Congress could direct the Attorney General on an emergency basis to figure out which drugs to classify as controlled substances. And United States v. O’Hagan, 521 U.S. 642, 117 S.Ct. 2199, 138 L.Ed.2d 724 (1997), saw nothing objectionable in a law authorizing the Securities and Exchange Commission to make rules combating securities fraud and to make violations of these rules crimes. If the Court allowed Congress to assign responsibility for defining crimes to the executive in those cases, what makes today’s case different?
The argument overlooks the reality that, if Congress wants to assign the executive branch discretion to define criminal conduct, it must speak “distinctly.” Grimaud, 220 U.S.. at 519, 31 S.Ct. 480; United States v. Eaton, 144 U.S. 677, 688, 12 S.Ct. 764, 36 L.Ed. 591 (1892). This clear-statement rule reinforces horizontal separation of powers in the same way that Gregory v. Ashcroft, 501 U.S. 452, 111 S.Ct. 2395, 115 L.Ed.2d 410 (1991), reinforces vertical separation of powers. It compels Congress to legislate deliberately and explicitly before departing from the Constitution’s traditional distribution of authority. Cases like Grimaud, Touby and O’Hagan respected this express-statement requirement, but the government’s theory flouts it. Under the government’s approach, courts could presume a congressional delegation of authority to create crimes whenever a criminal statute contains a gap. A presumption does not a clear statement make.
A related analogy to the Court’s federalism precedents fortifies the point. The *734Constitution sometimes allows Congress to upset federalism norms provided it legislates clearly. See Gregory, 501 U.S. at 460, 111 S.Ct. 2395. But it does not follow that Chevron allows agencies to upset federalism norms when Congress legislates ambiguously. See SWANCC, 531 U.S. at 172-73, 121 S.Ct. 675. In the same way, Congress may sometimes depart from separation-of-powers principles so long as it legislates clearly. But it does not follow that agencies may depart from separation-of-powers principles when Congress legislates ambiguously.
Quite apart from the clear-statement rule, the Constitution may well also require Congress to state more than an “intelligible principle” when leaving the definition of crime to the executive. The Supreme Court has suggested that “greater congressional specificity [may be] required in the criminal context.” Touby, 500 U.S. at 166, 111 S.Ct. 1752; see Yakus v. United States, 321 U.S. 414, 423-27, 64 S.Ct. 660, 88 L.Ed. 834 (1944). The laws at issue in Grimaud, Touby and O’Hagan honored this principle. But under the government’s approach, an agency could fill a gap in a criminal statute even where Congress provides no specific guidance about how to fill it.
The government separately relies heavily on a footnote in Babbitt v. Sweet Home Chapter of Communities for a Great Oregon. 515 U.S. 687, 704 n. 18, 115 S.Ct. 2407, 132 L.Ed.2d 597 (1995). Sweet Home arose under the Endangered Species Act, which made it an offense (subject to civil and criminal penalties) to “take” any endangered species. 16 U.S.C. § 1538(a)(1). The Interior Department issued a regulation interpreting this provision to prohibit “significant habitat modification or degradation” that kills or injures protected wildlife. 50 C.F.R. § 17.3 (1994). Before the agency could enforce this regulation, landowners challenged it on its face, claiming that it outstripped the agency’s statutory authority.
Citing Chevron, the Court gave the interpretation contained in the regulation “some degree of deference.” Sweet Home, 515 U.S. at 703, 115 S.Ct. 2407. The Court then dropped this footnote:
Respondents also argue that the rule of lenity should foreclose any deference to the Secretary’s interpretation ... because the statute includes criminal penalties .... We have applied the rule of lenity in a case raising a narrow question concerning the application of a statute that contains criminal sanctions to a specific factual dispute ... where no regulation was present. See United States v. Thompson/Center Arms Co., 504 U.S. 505, 517-18 & n. 9 [112 S.Ct. 2102, 119 L.Ed.2d 308] (1992). We have never suggested that the rule of lenity should provide the standard for reviewing facial challenges to administrative regulations whenever the governing statute authorizes criminal enforcement. Even if there exist regulations whose interpretations of statutory criminal penalties provide such inadequate notice of potential liability as to offend the rule of lenity, the [present] regulation, which has existed for two decades and gives a fair warning of its consequences, cannot be one of them.
Id. at 704 n. 18, 115 S.Ct. 2407. As the government reads it, this passage definitively holds that Chevron deference defeats the rule of lenity.
That is a lot to ask of a footnote, more it seems to me than these four sentences can reasonably demand. Note first of all that the government’s reading eclipses the just-mentioned Grimaud/Eaton line of cases, which hold that, if Congress wants to assign responsibility for crime definition to the executive, it must speak clearly. No *735one thinks that Chevron-triggering ambiguity satisfies a clear-statement requirement. Did the Court mean to overrule these precedents in a footnote that does not even mention them? Not likely. And a case decided after Sweet Home expressly declines — in a footnote, no less — to decide how the rule of lenity and Chevron interact. See SWANCC, 531 U.S. at 174 n. 8, 121 S.Ct. 675. Why did the Court express reluctance to decide a question if, as the government claims, it had already decided it?
The answer is that Sweet Home’s footnote 18 lends itself to a narrower reading, one that preserves the clear-statement rule applicable in this setting and one that preserves the obligation of courts and agencies to respect the rule of lenity. The footnote merely acknowledges the possibility of a pre-enforcement facial challenge to an agency’s regulation — because the agency had no interpretive authority in the first place, because the agency failed to follow the procedures for promulgating the regulation or because the statute plainly forecloses the agency’s interpretation. Yet not one of these challenges depends on, or demands consideration of, the rule of lenity. Why else would the Court distinguish cases involving “specific factual dispute[s]” from cases “reviewing facial challenges”? What purpose could this distinction serve unless the Court meant to create a rule for facial challenges? Although the footnote mentions that the Interior Department’s two-decade-old regulation comports with one of the rule of lenity’s objectives (promoting fair notice), it says nothing about other regulations or the rule of lenity’s separation-of-powers objective (reinforcing that Congress, not courts or agencies, define crimes). Before accepting the government’s broad reading of the footnote, one would have expected the Court to say more before allowing agencies to trump a doctrine Chief Justice Marshall described as “perhaps not much less old than construction itself.” 5 Wheat, at 95.
Not only does the age of the rule counsel against sweeping it aside sotto voce in a footnote, but so does its growing significance in interpretive disputes about the meaning of criminal laws. The Court has all but abandoned the practice of interpreting criminal laws against defendants on the basis of legislative history. Compare, e.g., United States v. Santos, 553 U.S. 507, 513 n. 3, 128 S.Ct. 2020, 170 L.Ed.2d 912 (2008) (plurality opinion), with, e.g., Dixson v. United States, 465 U.S. 482, 491, 104 S.Ct. 1172, 79 L.Ed.2d 458 (1984); see United States v. R.L.C., 503 U.S. 291, 307-11, 112 S.Ct. 1329, 117 L.Ed.2d 559 (1992) (Scalia, J., concurring in the judgment). And it has found lenity-triggering ambiguity in criminal laws more readily of late than it did in the past. Compare, e.g., Santos, 553 U.S. at 513-14, 128 S.Ct. 2020, with, e.g., Moskal v. United States, 498 U.S. 103, 108, 111 S.Ct. 461, 112 L.Ed.2d 449 (1990). Meanwhile, deference has shrunk in reach. The Court has cabined the range of materials entitled to Chevron deference. See, e.g., Mead, 533 U.S. at 231, 121 S.Ct. 2164. And it has confirmed that Chevron does not permit an agency to trump other rules of interpretation. See, e.g., St. Cyr, 533 U.S. at 320 n. 45, 121 S.Ct. 2271. Lenity and Chevron thus look different now than they did when Sweet Home inscrutably footnoted their interaction.
If accepted, moreover, the government’s theory would mean that, in many of the Court’s criminal cases, criminal defendants were one agency interpretation away from being incarcerated. It would mean that ambiguity in the Fair Labor Standards Act would have allowed the Secretary of Labor to decide that an employer commits a new crime not just when he sets an employee’s salary too low but each week he underpays the employee. But see United States v. *736Universal C.I.T. Credit Corp., 344 U.S. 218, 73 S.Ct. 227, 97 L.Ed. 260 (1952). It would mean that ambiguity in the National Firearms Act would have allowed the Secretary of the Treasury to decide that a gun manufacturer commits a crime when it packages a pistol with a carbine-conversion kit. But see United States v. Thompson/Center Arms Co., 504 U.S. 505, 112 S.Ct. 2102, 119 L.Ed.2d 308 (1992). It would mean that ambiguity in the Immigration and Nationality Act would have allowed the Attorney General to decide that drunk driving is a crime of violence. But see Leocal v. Ashcroft, 543 U.S. 1, 125 S.Ct. 377, 160 L.Ed.2d 271 (2004). And it would mean that ambiguity in the Immigration and Nationality Act would have allowed the Attorney General to decide that first-time drug possession and social sharing of marijuana are drug-trafficking crimes. But see Carachuri-Rosendo v. Holder, 560 U.S. 563, 130 S.Ct. 2577, 177 L.Ed.2d 68 (2010); Moncrieffe v. Holder, — U.S.-, 133 S.Ct. 1678, 185 L.Ed.2d 727 (2013).
In the final analysis, the government’s theory gives the executive branch an implied share of the legislature’s power to define crimes. That is no small matter given “the growing power of the administrative state,” City of Arlington v. FCC, — U.S. -, 133 S.Ct. 1863, 1879, — L.Ed.2d - (2013) (Roberts, C.J., dissenting), and it is no small matter given the reality that Congress continues to “put[ ] forth an ever-increasing volume ... of criminal laws,” Sykes v. United States, - U.S. -, 131 S.Ct. 2267, 2288, 180 L.Ed.2d 60 (2011) (Scalia, J., dissenting). None of the Supreme Court’s decisions requires us to accept this theory; many stand in its way. Agencies, no less than courts, must honor the rule of lenity.