OPINION

ALICE M. BATCHELDER, Circuit Judge.

Like many states, Michigan has amended its Sex Offender Registration Act (SORA) on a number of occasions in recent years for the professed purpose of making Michigan communities safer and aiding law enforcement in the task of bringing recidivists to justice. Thus, what began in 1994 as a non-public registry maintained solely for law enforcement use, see Mich. Pub. Act 295, § 10 (1994), has grown into a byzantine code governing in minute detail the lives of the state's sex offenders, see Mich. Comp. Laws § 28.723, et seq. Over the first decade or so of SORA's existence, most of the changes centered on the role played by the registry itself. In 1999, for example, the legislature added the requirement that sex offenders register in person (either quarterly or annually, depending on the offense) and made the registry available online, providing the public with a list of all registered sex offenders' names, addresses, biometric data, and, since 2004, photographs. See Mich. Pub. Act. 85 §§ 5a(4), 8(2), 10(2)(3) (1999); Mich. Pub. Acts 237, 238 (2004). Michigan began taking a more aggressive tack in 2006, however, when it amended SORA to prohibit registrants (with a few exceptions, see Mich. Comp. Laws § 28.734-36) from living, working, or "loitering"1 within 1,000 feet of a school. See Mich. Pub. Acts 121, 127 (2005). In 2011, the legislature added the requirement that registrants be divided into three tiers, which ostensibly correlate to current dangerousness, but which are based, not on individual assessments, but solely on the crime of conviction. See Mich. Pub. Acts 17, 18 (2011). The 2011 amendments also require all registrants to appear in person "immediately" to update information such as new vehicles or "internet identifiers" (e.g., a new email account). See id. The 2006 and 2011 amendments apply retroactively to all who were required to register under SORA. See Mich. Pub. Act 46 (2006); Mich. Pub. Acts 17, 18 (2011). Violations carry heavy criminal penalties. See Mich. Comp. Laws § 28.729.

The Plaintiffs in this case — identified here only as five "John Does" and one "Mary Doe" — are registered "Tier III" sex offenders currently residing in Michigan. It is undisputed on appeal that SORA's 2006 and 2011 amendments apply to them retroactively. That law has had a significant impact on each of them that reaches far beyond the stigma of simply being identified as a sex offender on a public registry. As a result of the school zone restrictions, for example, many of the Plaintiffs have had trouble finding a home in which they can legally live or a job where they can legally work. These restrictions have also kept those Plaintiffs who have children (or grandchildren) from watching them participate in school plays or on school sports teams, and they have kept Plaintiffs from visiting public playgrounds with their children for fear of "loitering." Plaintiffs are also subject to the frequent inconvenience of reporting to law enforcement in person whenever they change residences, change employment, enroll (or un-enroll) as a student, change their name, register a new email address or other "internet identifier," wish to travel for more than seven days, or buy or begin to use a vehicle (or cease to own or use a vehicle). See Mich. Comp. Laws §§ 28.722(g), 725(1).

Plaintiffs sued Michigan Governor Richard Snyder and Colonel Kriste Etue, the director Michigan's state police (collectively, "Michigan"), challenging SORA's validity on a number of different grounds, including that portions of SORA are unconstitutionally vague, that its requirements should not be construed as creating strict liability offenses, that SORA violates the right to free speech guaranteed by the First Amendment, and that it violates the Fourteenth Amendment by imposing oppressive restrictions on Plaintiffs' ability to parent, work, and travel. Plaintiffs also contended that SORA's retroactive application to them — specifically, the retroactive application of the amendments that went into effect starting in 2006 or later — amounts to an Ex Post Facto punishment prohibited by the Constitution. See U.S. Const. art. I, § 10, cl. 1.

In a handful of opinions, including an opinion following from a Rule 52 bench trial, the district court concluded, among other things, that SORA was not an Ex Post Facto law and that most of its provisions did not violate the Constitution's guarantee of due process. It did conclude, however, that Plaintiffs were correct that some of SORA's provisions were unconstitutionally vague, that those who are required to register under that law cannot be held strictly liable for violating its requirements, and that its retroactive requirement that sex offenders register on-line aliases for life violated the First Amendment. Both sides filed timely appeals, which we have consolidated.

We begin our analysis with the Ex Post Facto issue. As is the case with many of the Constitution's guarantees — "due process of law," "the freedom of speech," "the right of the people to keep and bear arms" — the Ex Post Facto clause leaves unanswered foundational questions about the guarantee's scope and means of enforcement. The document itself provides simply that "No State shall ... pass any... ex post facto Law." U.S. Const. art. I § 10, cl. 1. As with the other guarantees, it is the courts that have done most of the work in expounding the legal meaning of this provision — indeed, the Ex Post Facto clause was one of the first, if not the first, such constitutional question to be exposited by the Supreme Court, when it issued its 1798 decision in Calder v. Bull, 3 U.S. 386, 3 Dall. 386, 1 L.Ed. 648 (1798). That case, consistent with what scholars have identified as the majority position at the time of the founding, held that the Constitution's ban on Ex Post Facto laws does not bar all retroactive lawmaking, but only retroactive punishment, a codification of what many in the founding generation believed to be a self-evident truth: nulle poena sine lege, no punishment without a law. See Calder, 3 U.S. at 388 (Opinion of Chase, J.) (explaining the Court's holding that the Ex Post Facto clause prohibits only retroactive punishment and opining that such punishment was so "contrary to the great first principles of the social compact [that it could not] be considered a rightful exercise of legislative authority" even if there were no provision in the Constitution prohibiting it); see also David F. Forte, Ex Post Facto, in The Heritage Guide to the Constitution, 203, 203-04 (David F. Forte & Matthew Spalding, eds. 2d ed. 2014).

This understanding has kept courts from interfering with state sovereignty in many cases, but it has also provided a powerful check on states when they have sought to punish socially disfavored persons without prior notice. As Chief Justice John Marshall explained in Fletcher v. Peck:

10 U.S. 87, 137-38, 6 Cranch 87, 3 L.Ed. 162 (1810). The guarantee is, as James Madison put it, a "constitutional bulwark in favour of personal security and private rights." The Federalist No. 44, at 232 (James Madison) (The Gideon ed., George W. Carey & James McClellan eds., Liberty Fund 2001). As these quotations suggest, moreover, the distinction between civil regulation and criminal punishment has never been woodenly applied. "[I]t is the effect, not the form, of the law that determines whether it is ex post facto." Weaver v. Graham, 450 U.S. 24, 31, 101 S.Ct. 960, 67 L.Ed.2d 17 (1981).

And while some have questioned the self-evidence of nulle poena sine lege, see, e.g., Calder, 3 U.S. at 399 (Opinion of Iredell, J.) ("The ideas of natural justice are regulated by no fixed standard: the ablest and the purest men have differed upon the subject."), and others have called into doubt the correctness of Calder's civil/criminal distinction, see, e.g., Eastern Enterprises v. Apfel, 524 U.S. 498, 539, 118 S.Ct. 2131, 141 L.Ed.2d 451 (1998) (Thomas, J., concurring) ("I would be willing to reconsider Calder and its progeny to determine whether a retroactive civil law that passes muster under our current Takings Clause jurisprudence is nonetheless unconstitutional under the Ex Post Facto Clause."), the test we must apply, as a lower court, is quite fixed: an ostensibly civil and regulatory law, such as SORA, does not violate the Ex Post Facto clause unless the plaintiff can show "by the clearest proof" that "what has been denominated a civil remedy" is, in fact, "a criminal penalty," Smith v. Doe, 538 U.S. 84, 92, 123 S.Ct. 1140, 155 L.Ed.2d 164 (2003) (citation and internal quotation marks omitted).

The Supreme Court's decision in Smith is particularly germane to this case. In Smith, the Court considered an Ex Post Facto challenge to Alaska's sex-offender registry law. Alaska's regime was more modest than SORA, but the two share some core provisions: sex offenders residing in Alaska had to submit to annual or quarterly registration (though not in person) and had to give the State updates for such things as moving, growing a beard, changing hair color, or getting a new car. Id. at 90-91, 123 S.Ct. 1140; 101, 123 S.Ct. 1140. Like Michigan under SORA, Alaska maintained a website that published the offenders' names, addresses, photos, physical descriptions, license numbers, places of employment, dates of birth, crimes of conviction, dates and places of conviction, and length of sentences, as well the offenders' compliance with the registration requirements. Id. at 91, 123 S.Ct. 1140.

The Court in Smith concluded that Alaska's law was civil, not criminal, employing a two-part test: (1) Did the legislature intend to impose punishment? And (2), if not, is the statutory scheme "`so punitive either in purpose or effect as to negate [the State's] intention' to deem it `civil.'" Id. at 92, 123 S.Ct. 1140 (quoting Kansas v. Hendricks, 521 U.S. 346, 361, 117 S.Ct. 2072, 138 L.Ed.2d 501 (1997)) (brackets in original).

With respect to the first question, SORA, like the Alaska statute, includes a statement of purpose that evinces no punitive intent:

Mich. Comp. Laws § 28.721a. And while Plaintiffs do point to some features that might suggest a punitive aim — e.g., SORA is triggered solely by criminal offenses and the registration requirement is recorded on the judgment; registration is handled by criminal justice agencies like the police; SORA imposes criminal sanctions; and it is codified in Chapter 28 of the Michigan Code, a chapter that deals with police-related laws — these are similar enough to the arguments rejected in Smith that we see no warrant for concluding that SORA's intent is punitive. See Smith, 538 U.S. at 95, 123 S.Ct. 1140.

We must therefore consider whether SORA's actual effects are punitive. Out of the seven factors — non-dispositive "guideposts" — that typically inform this inquiry, Smith identified five that are relevant in this kind of case:

Smith, 538 U.S. at 97, 123 S.Ct. 1140 (citing Kennedy v. Mendoza-Martinez, 372 U.S. 144, 168-69, 83 S.Ct. 554, 9 L.Ed.2d 644 (1963)). We consider each factor in turn.

History and Tradition. As amici law professors point out, though SORA has no direct ancestors in our history and traditions, its restrictions do meet the general, and widely accepted, definition of punishment offered by legal philosopher H.L.A. Hart: (1) it involves pain or other consequences typically considered unpleasant; (2) it follows from an offense against legal rules; (3) it applies to the actual (or supposed) offender; (4) it is intentionally administered by people other than the offender; and (5) it is imposed and administered by an authority constituted by a legal system against which the offense was committed. See H.L.A. Hart, Punishment and Responsibility 4-5 (1968).

More specifically, SORA resembles, in some respects at least, the ancient punishment of banishment. True, it does not prohibit the registrant from setting foot in the school zones, and it certainly doesn't make a registrant "dead in law [and] entirely cut off from society," which is how Blackstone described the banished. 1 William Blackstone, Commentaries *132. But its geographical restrictions are nevertheless very burdensome, especially in densely populated areas. Consider, for example, this map of Grand Rapids, Michigan, prepared by one of Plaintiff's expert witnesses:

"School safety zones" in the city of Grand Rapids

Sex Offenders are forced to tailor much of their lives around these school zones, and, as the record demonstrates, they often have great difficulty in finding a place where they may legally live or work. Some jobs that require traveling from jobsite to jobsite are rendered basically unavailable since work will surely take place within a school zone at some point.

SORA's requirements also resemble traditional shaming punishments. Unlike the law in Smith, which republished information that was already publically available, SORA ascribes and publishes tier classifications corresponding to the state's estimation of present dangerousness without providing for any individualized assessment. These designations are unappealable, and apply even to those whose offenses would not ordinarily be considered sex offenses. Doe #1, for example, is on the registry because of a non-sexual kidnapping offense arising out of a 1990 robbery of a McDonald's. In other cases, SORA discloses otherwise non-public information. Doe #2, for example, is on the registry because, in 1996, when he was eighteen years old, he pled guilty under Michigan's "Holmes Youthful Trainee Act," Mich. Comp. Laws § 762.11, a record-sealing statute for young offenders, to "Criminal Sexual Conduct III" for having sex with a fourteen-year-old girl with whom he had a romantic relationship. But for SORA's retroactive application to him, his criminal record would not be available to the public. Thus, unlike the statute in Smith, the ignominy under SORA flows not only from the past offense, but also from the statute itself.

Finally, SORA also resembles the punishment of parole/probation. In Smith, which involved nothing more than reporting requirements, the Court took seriously the claim that the Alaska statute resembled parole/probation, acknowledging that "[t]his argument has some force, but," concluding that it was ultimately dissimilar because, unlike parolees, "offenders subject to the Alaska statute are free to move where they wish and to live and work as other citizens, with no supervision." 538 U.S. at 101, 123 S.Ct. 1140. Under SORA, by contrast, registrants are subject to numerous restrictions on where they can live and work and, much like parolees, they must report in person, rather than by phone or mail. Failure to comply can be punished by imprisonment, not unlike a revocation of parole. And while the level of individual supervision is less than is typical of parole or probation, the basic mechanism and effects have a great deal in common. In fact, many of the plaintiffs have averred that SORA's requirements are more intrusive and more difficult to comply with than those they faced when on probation.

In sum, while SORA is not identical to any traditional punishments, it meets the general definition of punishment, has much in common with banishment and public shaming, and has a number of similarities to parole/probation. This factor thus weighs in Plaintiffs' favor.

Affirmative Disability or Restraint. As should be evident, SORA requires much more from registrants than did the statute in Smith. Most significant is its regulation of where registrants may live, work, and "loiter." As discussed above, these restrictions put significant restraints on how registrants may live their lives. Further, as also mentioned above, registrants must appear in person, both initially and for updates, and, if they are "Tier III" offenders, they must do so for life. These are direct restraints on personal conduct.

Michigan points out, however, that these restraints are not physical in nature and contends that the actual effects are therefore "minor and indirect" like those in the statute considered in Smith, 538 U.S. at 100, 123 S.Ct. 1140. But surely something is not "minor and indirect" just because no one is actually being lugged off in cold irons bound. Indeed, those irons are always in the background since failure to comply with these restrictions carries with it the threat of serious punishment, including imprisonment. These restraints are greater than those imposed by the Alaska statute by an order of magnitude. Cf. Smith, 538 U.S. at 101, 123 S.Ct. 1140 (noting, for example, that "[t]he Alaska statute, on its face, does not require these updates to be made in person").

Michigan has a stronger point in noting that SORA's restrictions are in some ways not as severe as complete occupation-disbarment, which has been held to be non-punitive. Id. at 100, 123 S.Ct. 1140; see also De Veau v. Braisted, 363 U.S. 144, 80 S.Ct. 1146, 4 L.Ed.2d 1109 (1960) (forbidding work as a union official); Hawker v. New York, 170 U.S. 189, 18 S.Ct. 573, 42 L.Ed. 1002 (1898) (revocation of a medical license). But no disbarment case we are aware of has confronted a law with such sweeping conditions or approved of disbarment without some nexus between the regulatory purpose and the job at issue. SORA's restrictions are again far more onerous than those considered in Smith. And this factor too therefore weighs in Plaintiffs' favor.

Traditional Aims of Punishment. SORA advances all the traditional aims of punishment: incapacitation, retribution, and specific and general deterrence. Its very goal is incapacitation insofar as it seeks to keep sex offenders away from opportunities to reoffend. It is retributive in that it looks back at the offense (and nothing else) in imposing its restrictions, and it marks registrants as ones who cannot be fully admitted into the community. Further, as discussed below, it does so in ways that relate only tenuously to legitimate, non-punitive purposes. Finally, its professed purpose is to deter recidivism (though, as discussed below, it does not in fact appear to do so), and it doubtless serves the purpose of general deterrence. See J.J. Prescott & Jonah E. Rockoff, Do Sex offender Registration and Notification Laws Affect Criminal Behavior?, 54 J.L. & Econ. 161 (2011).

Of course, many of these goals can also rightly be described as civil and regulatory. See Smith, 538 U.S. at 102, 123 S.Ct. 1140 ("Any number of governmental programs might deter crime without imposing punishment. To hold that the mere presence of a deterrent purpose renders such sanctions criminal would severely undermine the Government's ability to engage in effective regulation." (internal quotation marks and citation omitted)). And we accordingly give this factor little weight.

Rational Relation to a Non-Punitive Purpose. "The Act's rational connection to a nonpunitive purpose is a `[m]ost significant' factor in our determination that the statute's effects are not punitive." Id. (quoting United States v. Ursery, 518 U.S. 267, 290, 116 S.Ct. 2135, 135 L.Ed.2d 549 (1996)) (brackets in original). As in Smith, the legislative reasoning behind SORA is readily discernible: recidivism rates of sex offenders, according to both the Michigan legislature and Smith, are "frightening and high"; informing the public of sex offenders' addresses, photos, tier rankings, etc. provides a mechanism to keep tabs on them with a view to preventing some of the most disturbing and destructive criminal activity; and school zones keep sex offenders away from the most vulnerable.

Intuitive as some may find this, the record before us provides scant support for the proposition that SORA in fact accomplishes its professed goals. The record below gives a thorough accounting of the significant doubt cast by recent empirical studies on the pronouncement in Smith that "[t]he risk of recidivism posed by sex offenders is `frightening and high.'" 538 U.S. at 103, 123 S.Ct. 1140 (quoting McKune v. Lile, 536 U.S. 24, 34, 122 S.Ct. 2017, 153 L.Ed.2d 47 (2002)). One study suggests that sex offenders (a category that includes a great diversity of criminals, not just pedophiles) are actually less likely to recidivate than other sorts of criminals. See Lawrence A. Greenfield, Recidivism of Sex Offenders Released from Prison in 1994 (2003). Even more troubling is evidence in the record supporting a finding that offense-based public registration has, at best, no impact on recidivism. [R. 90 at 3846-49]. In fact, one statistical analysis in the record concluded that laws such as SORA actually increase the risk of recidivism, probably because they exacerbate risk factors for recidivism by making it hard for registrants to get and keep a job, find housing, and reintegrate into their communities. See Prescott & Rockoff, supra at 161. Tellingly, nothing the parties have pointed to in the record suggests that the residential restrictions have any beneficial effect on recidivism rates. And while it is intuitive to think that at least some sex offenders — e.g., the stereotypical playground-watching pedophile — should be kept away from schools, the statute makes no provision for individualized assessments of proclivities or dangerousness, even though the danger to children posed by some — e.g., Doe #1, who never committed a sexual offense — is doubtless far less than that posed by a serial child molester.

Excessiveness. Further, while the statute's efficacy is at best unclear, its negative effects are plain on the law's face. As explained above, SORA puts significant restrictions on where registrants can live, work, and "loiter," but the parties point to no evidence in the record that the difficulties the statute imposes on registrants are counterbalanced by any positive effects. Indeed, Michigan has never analyzed recidivism rates despite having the data to do so. [R. 90 at 3768-69]. The requirement that registrants make frequent, in-person appearances before law enforcement, moreover, appears to have no relationship to public safety at all. The punitive effects of these blanket restrictions thus far exceed even a generous assessment of their salutary effects.

So, is SORA's actual effect punitive? Many states confronting similar laws have said "yes." See, e.g., Doe v. State, 167 N.H. 382, 111 A.3d 1077, 1100 (2015); State v. Letalien, 985 A.2d 4, 26 (Me. 2009); Starkey v. Oklahoma Dep't of Corr., 305 P.3d 1004 (Okla. 2013); Commonwealth v. Baker, 295 S.W.3d 437 (Ky. 2009); Doe v. State, 189 P.3d 999, 1017 (Alaska 2008). And we agree. In reaching this conclusion, we are mindful that, as Smith makes clear, states are free to pass retroactive sex-offender registry laws and that those challenging an ostensibly non-punitive civil law must show by the "clearest proof" that the statute in fact inflicts punishment. But difficult is not the same as impossible. Nor should Smith be understood as writing a blank check to states to do whatever they please in this arena.

A regulatory regime that severely restricts where people can live, work, and "loiter," that categorizes them into tiers ostensibly corresponding to present dangerousness without any individualized assessment thereof, and that requires time-consuming and cumbersome in-person reporting, all supported by — at best — scant evidence that such restrictions serve the professed purpose of keeping Michigan communities safe, is something altogether different from and more troubling than Alaska's first-generation registry law. SORA brands registrants as moral lepers solely on the basis of a prior conviction. It consigns them to years, if not a lifetime, of existence on the margins, not only of society, but often, as the record in this case makes painfully evident, from their own families, with whom, due to school zone restrictions, they may not even live. It directly regulates where registrants may go in their daily lives and compels them to interrupt those lives with great frequency in order to appear in person before law enforcement to report even minor changes to their information.

We conclude that Michigan's SORA imposes punishment. And while many (certainly not all) sex offenses involve abominable, almost unspeakable, conduct that deserves severe legal penalties, punishment may never be retroactively imposed or increased. Indeed, the fact that sex offenders are so widely feared and disdained by the general public implicates the core counter-majoritarian principle embodied in the Ex Post Facto clause. As the founders rightly perceived, as dangerous as it may be not to punish someone, it is far more dangerous to permit the government under guise of civil regulation to punish people without prior notice. Such lawmaking has "been, in all ages, [a] favorite and most formidable instrument[] of tyranny." The Federalist No. 84, supra at 444 (Alexander Hamilton). It is, as Justice Chase argued, incompatible with both the words of the Constitution and the underlying first principles of "our free republican governments." Calder, 3 U.S. at 388-89; accord The Federalist No. 44, supra at 232 (James Madison) ("[E]x post facto laws ... are contrary to the first principles of the social compact, and to every principle of sound legislation."). The retroactive application of SORA's 2006 and 2011 amendments to Plaintiffs is unconstitutional, and it must therefore cease.

As we have explained, this case involves far more than an Ex Post Facto challenge. And as the district court's detailed opinions make evident, Plaintiffs' arguments on these other issues are far from frivolous and involve matters of great public importance. These questions, however, will have to wait for another day because none of the contested provisions may now be applied to the plaintiffs in this lawsuit, and anything we would say on those other matters would be dicta. We therefore reverse the district court's decision that SORA is not an Ex Post Facto law and remand for entry of judgment consistent with this opinion.