Wilfred Holmes v. Christopher Christie

PRECEDENTIAL UNITED STATES COURT OF APPEALS FOR THE THIRD CIRCUIT ____________ No. 19-1089 ____________ WILFRED LEE HOLMES, Appellant v. N.J. GOV. CHRISTOPHER J. CHRISTIE, both individually and in his official capacity as the Governor of the State of New Jersey; DAVID W. THOMAS, both individually and in his official capacity as the executive director of the N.J. State Parole Board; JAMES T. PLOUSIS, both individually and in his official capacity as chairman of the N.J. State Parole Board; SAMUEL PLUMERI, JR., both individually and in his official capacity as vice-chairman of the N.J. State Parole Board; STUART RABNER, both individually and in his official capacity as the chief justice of the Supreme Court; CARMEN MESSANO, both individually and in her official capacity as the acting presiding judge for the administration for the Superior Court of New Jersey, Appellate Division; MARGARET M. HAYDEN; JOHN R. TASSINI, both individually and in their official capacities as judges of the Superior Court, Appellate Division ____________ On Appeal from the United States District Court for the District of New Jersey (D.N.J. No. 2:16-cv-1434) District Judge: Honorable Esther Salas ____________ Argued March 16, 2021 Before: KRAUSE, PHIPPS, and FUENTES, Circuit Judges (Opinion Filed: September 22, 2021) Julie Michalski Steptoe & Johnson 227 West Monroe Street Suite 4700 Chicago, IL 60606 Steven Reed Jessica I. Rothschild Mark C. Savignac [Argued] Steptoe & Johnson 1330 Connecticut Avenue, N.W. Washington, DC 20036 Pro Bono Counsel for Appellant Deborah A. Hay [Argued] Christopher C. Josephson Office of the Attorney General of New Jersey 25 Market Street Richard J. Hughes Justice Complex 8th Floor, West Wing Trenton, NJ 08625 Counsel for Appellees ____________ OPINION OF THE COURT ____________ KRAUSE, Circuit Judge. The Ex Post Facto Clause prevents the government from increasing a prisoner’s punishment retroactively. This case requires us to decide whether the Clause permits New Jersey to retroactively enforce certain parole rules. To answer that question, we look to the rules’ “practical effect” on each inmate’s chances of receiving early release. Richardson v. Pa. Bd. of Prob. & Parole, 423 F.3d 282, 290 (3d Cir. 2005). For 2 many prisoners, no doubt, the rules present at most a remote risk to their parole prospects. For the Appellant here, however, the change plausibly produced a significant risk of prolonging his time behind bars. Thus, we vacate the District Court’s dismissal order, reinstate Appellant’s ex post facto claim, and remand for discovery. I. Background A. New Jersey’s Parole System Before turning to the facts of this case, we introduce New Jersey’s parole system. Since its inception, that system has featured two types of parole hearings: initial hearings and successive hearings. When a prisoner first becomes eligible for release, New Jersey’s Parole Board holds a hearing, decides whether to grant parole, and, if it declines to do so, sets a date to revisit its decision. See N.J. Stat. Ann. § 30:4-123.53(a) (2011). In the course of these initial hearings, the Board may consult any information it deems relevant, including an inmate’s criminal history. See id. Before 1997, however, a different set of evidentiary rules governed successive parole hearings. Under those rules, the Board could not consider old information,1 see id. § 30:4- 123.56(c) (1996), and instead based successive parole decisions “strictly on information developed since the previous denial of parole,” Assembly Law and Public Safety Committee, Statement to Assembly Bill No. 21 (Mar. 3, 1997). In practice, this prevented the Board from taking account of inmates’ criminal history—often the most damaging aspect of their records—after the initial hearing. The change wrought in 1997 had its roots in the early 1990s when many states moved to recalibrate their parole regimes. See, e.g., Mickens-Thomas v. Vaughn, 321 F.3d 374, 380 (3d Cir. 2003) (describing the impetus behind contemporaneous changes in Pennsylvania’s parole law). Not 1 New Jersey’s 1948 Parole Act governs Holmes’s case. Although New Jersey repealed that statute in enacting its 1979 Parole Act, “the standards of the 1979 and 1948 Acts are identical with respect to inmates convicted before 1979.” Royster v. Fauver, 775 F.2d 527, 535 (3d Cir. 1985). 3 content to sit on the sidelines, New Jersey’s then-Governor appointed a Commission to study the state’s parole system and propose reforms. The history from that point on is described in detail in Trantino v. New Jersey State Parole Board—a seminal state court opinion in which the Superior Court’s Appellate Division upheld that change under the Ex Post Facto Clause as merely “procedural” and not “substantive.” 752 A.2d 761, 780–82 (N.J. Super. App. Div. 2000). According to the Appellate Division, the purpose of the Commission was to “recommend legislation that would ‘enlarge the discretion of the Board to deny parole,’” and the Commission’s final report documented the practical effects of the rule against considering old information in successive hearings. Id. at 780 (quoting James Holzapfel, et al., Final Report of the Study Commission on Parole (Dec. 1996), [hereinafter, Final Report]2). Among those effects were that “‘the Board [wa]s effectively required to grant parole, even though the inmate may not be rehabilitated.’” Id. (quoting Final Report at *21). And because the Commission ranked the rule as “one of the most significant and inappropriate limitations that existing law place[d] on the Board’s discretion,” it urged New Jersey’s legislature to relax the rule and allow the Parole Board to examine “all relevant information” at every hearing. Id. (internal quotation marks omitted) (quoting Final Report at *21–22). B. The 1997 Amendments Just a few months after the Commission released its Final Report, the New Jersey legislature implemented its recommendations in the 1997 Amendments to the Parole Act. See 1997 N.J. Sess. Law Serv. ch. 213. Two of those amendments undergird this appeal: • The All-Information Provision: Consistent with the Commission’s recommendation, the Amendments eliminated the prohibition against reviewing old information. Compare N.J. Stat. Ann. § 30:4-123.56(c) 2 The Final Report is available at: https://dspace.njstatelib.org/xmlui/bitstream/handle/10929/18 629/P9591996a.pdf. 4 (2011), with N.J. Stat. Ann. § 30:4-123.56(c) (1996). Under the new regime, the Board enjoys free rein to revisit an inmate’s criminal history during successive hearings. • The Risk-Assessment Requirement: The Amendments also instructed the Board to prepare an “objective risk assessment” before every parole hearing, including successive hearings. N.J. Stat. Ann. § 30:4-123.52(e) (2001). This assessment must incorporate old information—including an inmate’s “educational and employment history” and “family and marital history”— along with any other “static and dynamic factors which may assist the [B]oard.” Id. Since 1997, the Board has applied these changes to all prisoners, including those convicted before the Amendments came into force.3 See Trantino, 752 A.2d at 781. C. This Lawsuit Appellant Wilfred Lee Holmes is no stranger to New Jersey’s parole system.4 When Holmes was on parole in the early 1970s, he killed two acquaintances, carried out the 3 The Amendments also adjusted the standard governing parole requests. Before 1997, the Board could deny parole only if “a preponderance of the evidence” showed “a substantial likelihood that the inmate w[ould] commit a crime . . . if released.” N.J. Stat. Ann. § 30:4-123.56(c) (1996). Under the Amendments, however, the Board may refuse release whenever “a preponderance of the evidence [indicates] that the inmate has failed to cooperate in his or her own rehabilitation or that there is a reasonable expectation that the inmate will violate conditions of parole.” N.J. Stat. Ann. § 30:4-123.56(c) (2011). This change is not at issue here. 4 Because Holmes brought this appeal pro se, we invited Steptoe & Johnson to serve as pro bono counsel. We express our gratitude to Julie Michalski, Steven Reed, Jessica I. Rothschild, Mark C. Savignac, and their firm for accepting this matter pro bono, and we commend them for their excellent briefing and argument. Lawyers who act pro bono fulfill the highest service that members of the bar can offer to indigent parties and to the legal profession. 5 execution-style murder of a 69-year-old, and wounded a police officer who tried to arrest him. State courts subsequently convicted Holmes of multiple homicides and sentenced him to life in prison with the possibility of parole. Forty-eight years and several parole hearings later, Holmes remains behind bars. At the initial hearing, performed in 2001, the Board refused to release Holmes and scheduled a follow-up hearing for about a decade later. Then, in 2012, the Board held the hearing that is the subject of this appeal. To announce the results of that hearing, the Board issued a detailed written statement. For the most part, the statement examines evidence the pre-1997 rules would have excluded. It probes Holmes’s past parole violations, highlights the homicides that led to his life sentences, and scrutinizes the shootout that preceded his arrest. Aside from analyzing this old information, the statement also discusses Holmes’s 2012 interview with the Board, and his unblemished disciplinary record since his initial parole hearing. Without spelling out how much weight it placed on each of these factors, the Board rejected Holmes’s request for release. Convinced that the Board “should have considered only ‘new’ information,” Holmes implored New Jersey’s Appellate Division to vacate the Board’s decision on ex post facto grounds. Holmes v. N.J. State Parole Bd. (“Holmes I”), No. A- 1315-13T2, 2015 WL 4544689, at *7 (N.J. Super. App. Div. July 29, 2015). But relying on Trantino, the Appellate Division rebuffed Holmes’s claim and upheld the Board’s decision in full. Holmes then submitted a pro se complaint in federal 5 court. Though the complaint advanced at least a dozen different claims, the District Court focused on two that are at issue here. Holmes v. Christie (“Holmes II”), No. 16 Civ. 1434 (ES) (MAH), 2018 WL 6522922, at *1 (D.N.J. Dec. 12, 2018). The first challenged the Board’s retroactive application of the 1997 Amendments as contrary to the Ex Post Facto Clause. It is undisputed that Holmes’s claims may be raised in 5 a civil rights complaint under 42 U.S.C. § 1983, rather than in a habeas petition. See Wilkinson v. Dotson, 544 U.S. 74, 81– 82 (2005). 6 J.A. at 34. In his complaint, Holmes explicitly cited the Appellate Division’s decision in Trantino as “not[ing]” that the “express intent of the [1997] [A]mendment[s]” was to make it more difficult for prospective parolees to earn parole. J.A. at 30. Having clearly understood the nature of the claims Holmes was raising, the District Court reviewed the relevant case law and identified that New Jersey courts had already concluded that Trantino foreclosed Holmes’s ex post facto claim. Holmes II, 2018 WL 6522922, at *2 (quoting Holmes I, 2015 WL 4544689, at *7 (quoting Trantino, 752 A.2d at 681–82)). The second criticized the Board’s approach as inconsistent with the Due Process Clause. J.A. at 35. To right these alleged wrongs, Holmes requested that the Board hold a new parole hearing without examining old information. But when the Government moved to dismiss Holmes’s claims, the District Court assented.6 Notwithstanding the Appellate Division’s acknowledgement that the Board was “effectively required to grant parole” prior to the 1997 Amendments, Trantino, 752 A.2d at 780, as a practical matter the District Court rejected the ex post facto claim, reasoning that the “Board’s consideration of factors [related to] recidivism was consistent with the goals of either version of the [New Jersey] statute,” Holmes II, 2018 WL 6522922, at *6. In denying the due process claim, the Court observed that Holmes had received all procedural protections the Constitution requires. This timely appeal followed. II. Ex Post Facto Claim7 The central question presented is whether the Board’s decision conflicts with the Ex Post Facto Clause. Answering 6 The complaint names numerous New Jersey officials as Defendants. For the sake of simplicity, we refer to them collectively as the “Government.” 7 The District Court retained jurisdiction under 28 U.S.C. § 1331, and we wield jurisdiction under 28 U.S.C. § 1291. We exercise plenary review over a district court’s dismissal of claims under Rule 12(b)(6), see Geness v. Cox, 7 that question takes us on a journey with three stages. We first familiarize ourselves with the constitutional landmarks that guide our analysis. Following those landmarks, we find that Holmes’s claim merits discovery. And with our own sojourn complete, we chart where the Government’s counterarguments stray off course. A. Constitutional Landmarks Before setting out, we sketch the key features of what for many represents an unfamiliar legal landscape. Though the Ex Post Facto Clause rarely appears in casebooks or civics classrooms, the Framers ranked it among the Constitution’s most fundamental guarantees. See The Federalist No. 44, at 282 (C. Rossiter ed. 1961) (James Madison); id. No. 84, at 511 (Alexander Hamilton). The Clause continues to serve vital purposes today. It prohibits legislatures from “enacting arbitrary and vindictive” laws that target disfavored groups. Miller v. Florida, 482 U.S. 423, 429–30 (1987), abrogated on other grounds by Cal. Dep’t of Corr. v. Morales, 514 U.S. 499, 506–07 n.3 (1995); see also Peugh v. United States, 569 U.S. 530, 541 n.4 (2013). It promotes the separation of powers “by confining the legislature to penal decisions with prospective effect and the judiciary and executive to applications of existing penal law.” Weaver v. Graham, 450 U.S. 24, 29 n.10 (1981). And it provides citizens with “fair warning” as to a crime’s “effective sentence.” Id. at 28, 32. To achieve these ends, the Clause extends to some “changes in laws governing parole of prisoners.” Garner v. Jones, 529 U.S. 244, 250 (2000). At the Founding, long prison sentences were unusual, and parole was almost unknown. See Daniel S. Medwed, The Innocent Prisoner’s Dilemma: Consequences of Failing to Admit Guilt at Parole Hearings, 93 IOWA L. REV. 491, 498 (2008); Will Tress, Unintended Consequences: Defining Felony in the Early American Republic, 57 CLEV. ST. L. REV. 461, 468–70 (2009). In time, many states embraced parole regimes, and “eligibility for reduced imprisonment” emerged as a “significant factor” in a 902 F.3d 344, 353–54 (3d Cir. 2018), accepting the complaint’s factual allegations as true and construing them in the light most favorable to the non-moving party, see Weimer v. County of Fayette, 972 F.3d 177, 180 (3d Cir. 2020). 8 defendant’s punishment. Lynce v. Mathis, 519 U.S. 433, 445– 46 (1997) (quoting Weaver v. Graham, 450 U.S. 24, 32 (1981)). Thus, “an offender, prior to his conviction and sentencing, is entitled to know not only his maximum possible punishment, but also his or her chances of receiving early release.” Mickens-Thomas, 321 F.3d at 392. But the same logic that supports extending the Clause to parole also limits its reach. When a parole rule produces a “significant” risk of increasing a plaintiff’s time behind bars, retroactively applying the rule frustrates fair notice, and thus thwarts the Clause. Morales, 514 U.S. at 508. When a “minor” change presents only a “remote risk of impact on a prisoner’s expected term of confinement,” however, the Clause’s purposes remain undisturbed. Id. A limit on “the hours that prisoners may use the prison law library,” for example, poses no ex post facto problem. Id. at 508–09. The “controlling inquiry,” then, is whether a challenged rule “creates a significant risk of prolonging [the plaintiff’s] incarceration.” Garner, 529 U.S. at 250–51. A plaintiff can satisfy this risk-based standard in two ways. In some cases, the requisite risk is “inherent” in a new rule, so that the government contradicts the Clause whenever it enforces the rule retroactively. Id. at 251. In other cases, though, “the rule [will] not by its own terms show a significant risk.” Id. at 255. The question then becomes whether a sufficient risk arises from the rule’s “practical implementation.” Id. Which of these paths a plaintiff pursues carries important consequences. The first path presents a question of law courts can answer at the pleading stage; the second implicates a fact-intensive inquiry that may require discovery to resolve. Garner, 529 U.S. at 256 (remanding to consider discovery because “[t]he record before the Court of Appeals contained little information bearing on the level of risk created by the change in law.”); Richardson, 423 F.3d at 291. Few successful parole challenges follow the first path. What defines most parole regimes—including New Jersey’s— is that release decisions depend on “individualized[,] discretionary appraisals.” Perry v. N.J. State Parole Bd., 208 A.3d 439, 443 (N.J. Super. App. Div. 2019) (internal quotation marks and citation omitted). That means a parole rule’s 9 “practical effect” usually turns not on the rule’s terms, but on how those terms are implemented. Richardson, 423 F.3d at 290. This point is best illustrated by a pair of Supreme Court cases. The more typical case, Garner, centered on a change in the frequency of parole hearings. See 529 U.S. at 247. Before the change, Georgia’s parole board considered each prisoner’s case every three years; after the change, the board could wait up to eight years between reviews. Garner, 529 U.S. at 254. Rather than condemning the challenged rule as an “inherent” ex post facto violation, the Court recognized that the rule’s constitutionality hinged on the board’s “actual practices.” Id. at 251, 256. Imagine, for instance, that the board exercised its discretion to “expedite[ ] parole reviews in the event of a change in [a prisoner’s] circumstance[s].” Id. at 254 (internal quotation marks omitted). In that case, the rule would generate “only the most speculative and attenuated possibility” of prolonged imprisonment. Id. at 251 (internal quotation marks omitted). So, the Court concluded, Garner’s ex post facto claim rose or fell based on the rule’s “practical implementation.” Id. at 255. While Garner represents the norm, Lynce represents the exception. Under the early-release program Lynce reviewed, Florida prisoners accrued release credits whenever “the population of the state prison system exceeded predetermined levels.” 519 U.S. at 435. Once the credits “resulted in [certain] prisoners’ release from custody,” however, Florida cancelled the program and re-arrested those who benefited from it. Id. In these exceptional circumstances, the Court saw no need to look beyond the new rule’s terms. Unlike a change in a discretionary parole regime, the Court explained, cancelling the credits “unquestionably disadvantaged [the prisoners] because it resulted in [their] rearrest and prolonged [their] imprisonment.” Id. at 446–47. Read together, Garner and Lynce stand for a simple proposition: A rule’s terms establish an ex post facto violation only if they leave a parole board with little or no discretion. Otherwise, a plaintiff must show that the rule’s implementation presents a significant risk. With the risk-based standard as a 10 compass, and with the two paths for satisfying that standard as a map, we advance to the next stage of our analytical inquiry. B. Holmes’s Claim In challenging the Board’s parole decision, Holmes launches a multi-pronged assault on the retroactive application of the 1997 Amendments. His main thrust targets the all- information provision. He also levels more cursory attacks on the risk-assessment requirement. We address each argument in turn. 1. The All-Information Provision Holmes focuses most of his firepower on the all- information provision. First, he insists that the provision’s terms establish the requisite risk. Second, he argues that the Board implemented the provision in a way that plausibly conflicts with the Clause. As we explain, the first theory founders, but the second succeeds. a) Terms Our initial inquiry is whether the all-information provision’s terms produce a significant risk. In principle, the provision empowers the Board to examine old information during successive parole hearings. See N.J. Stat. Ann. § 30:4- 123.56 (2011). In practice, it enables the Board to entertain factors—especially an inmate’s criminal history—that often militate against release. Thus, common sense suggests the provision will sometimes prompt the Board to deny parole when it might otherwise have granted release. But whether that risk materialized in Holmes’s case depends on how the Board implements the all-information provision. Perhaps the Board continues its past practice of treating new information as dispositive. Perhaps the Board prioritizes old information that helps prisoners, such as family or educational history.8 Or perhaps Holmes has committed 8 Because the 1997 Amendment requires release “unless” evidence before the Board justifies continued confinement, Holmes maintains that the Board can consult 11 new disciplinary infractions that the Board views as foreclosing release, no matter what old information it considers. We cannot rule out these and other possibilities without reviewing “at least some evidence,” Richardson, 423 F.3d at 293, as to “the manner in which [the Board] is exercising its discretion,” Garner, 529 U.S. at 256. The upshot is that this review of the all-information provision’s terms leads to a dead-end. Whether the provision’s retroactive application passes constitutional muster depends not on its terms, but on how the Board implements them. b) Implementation With Holmes’s initial theory out of the picture, we come to the crux of this case. To prevail here, Holmes must establish that the Board implemented the all-information provision in a way that created a significant risk of prolonging his imprisonment. See id. at 255. This is a “fact-intensive inquiry,” Richardson, 423 F.3d at 291, but to survive a motion to dismiss, he “need only show that his ex post facto claim— like any other claim—is ‘plausible,’” Daniel v. Fulwood, 766 F.3d 57, 61–62 (D.C. Cir. 2014) (quoting Ashcroft v. Iqbal, 556 U.S. 662, 678 (2009)). Holmes easily clears that hurdle, especially when his complaint is “liberally read under the relaxed standards applicable to a pro se complaint.” Tunnell v. Wiley, 514 F.2d 971, 974 (3d Cir. 1975). To demonstrate that the provision harmed him, he marshals two types of support: The Board’s written statement and its historical practices. The statement suggests that old information influenced the Board’s decision “in [Holmes’s] case.” Richardson, 423 F.3d at 293. And as a historical matter, he plausibly alleges that “similarly situated only old information that harms prisoners. N.J. Stat. Ann. § 30:4-123.56(c) (2011). But Holmes’s conclusion does not follow from his premise. The word “unless” establishes that the aggregate evidence before the Board must weigh against release, not that every individual piece of evidence must do so. Indeed, the Board’s own regulations authorize the review of some categories of old information—such as inmates’ military service—that normally favor release. See N.J. Admin. Code § 10A:71-3.11(b)(18). 12 inmates” tended to win release before the all-information provision came into effect. Newman v. Beard, 617 F.3d 775,786 (3d Cir. 2010). At the pleading stage, then, Holmes’s claim survives dismissal. i. The Board’s Written Statement To explain why it refused to release Holmes, the Board prepared a ten-page written statement.9 The statement does not spell out how much weight the Board placed on old information. But a reasonable reading of the statement’s structure and substance reveals that Holmes’s criminal history played an important part in the result, plausibly alleging that consideration of that history created a significant risk of prolonging his imprisonment. We start with structure. Three structural features signal that old information influenced the Board’s inquiry. First, the bulk of the statement—more than five of its ten pages—trains on Holmes’s criminal history. Second, the statement does not designate that history as background, but instead includes it in the same section as its other analysis. Finally, of six headers summarizing why the Board refused to release Holmes, five rest on old information. For example, the statement underscores the “increasingly more serious” nature of Holmes’s crimes, his track record of “commit[ing] new offenses” while on parole, and the failure of prior incarcerations to “deter [his] criminal behavior.” J.A. at 46– 48. In combination, these structural features make it reasonable to infer that Holmes’s criminal history contributed to the Board’s conclusion. That inference finds support in the decision’s substance. Recall that New Jersey’s Parole Act tasks the Board with predicting whether Holmes “will commit a crime . . . if 9 The statement purports to summarize the Board’s “reasons for establishing a future parole eligibility date outside of the administrative guidelines,” J.A. at 45, but the parties agree that it also explains the parole denial. 13 released.”10 N.J. Stat. Ann. § 30:4-123.56(c) (1996). If someone set out to construct a criminal history that portends future crimes, it might look a lot like Holmes’s. First in 1963, then in 1965, and again in 1968, 1969, 1970, 1971, and 1972, Holmes won early release, only to violate the conditions of his parole. And each time he was freed, he committed more serious offenses, culminating in a series of homicides. It is plausible that this pattern of escalating parole violations prompted the Board to conclude that Holmes would commit yet another crime if released. In an ordinary case, a parole board’s careful scrutiny of a prisoner’s past conduct upon early release would provide a reason to uphold the board’s decision, not to question its constitutionality—particularly where the board’s stated goal is to identify a “substantial likelihood that the inmate will commit a crime . . . if released.” N.J. Stat. Ann. § 30:4-123.56(c) (1996). But at the time of Holmes’s crimes, New Jersey forbade its Board from examining criminal history during successive parole hearings. See id. The Board’s detailed discussion of Holmes’s history demonstrates that it retroactively applied the 1997 Amendments to Holmes in denying him parole, plausibly substantiating the risk that the all-information provision had the “practical effect” of extending his imprisonment. Richardson, 423 F.3d at 289. ii. The Board’s Historical Practices Our review of the Board’s written decision establishes that the retroactive consideration of Holmes’s past conduct plausibly “create[d] a significant risk of prolonging [his] incarceration.” Garner, 529 U.S. at 250–51. But rather than rely on the Board’s statement alone, Holmes also argues that similarly-situated prisoners tended to win release before 1997. That conclusion is bolstered by Trantino and the Final Report. Recall that the purpose of the Commission was to “enlarge the discretion of the Board to deny parole,” and it recommended eliminating the old parole rules specifically because those rules “effectively required [the Board] to grant parole” unless an inmate had committed “institutional infractions . . . since his or 10 It is undisputed that this standard, excerpted from the pre-1997 rules, governed the Board’s decision in Holmes’s case. See Holmes I, 2015 WL 4544689, at *4. 14 her last review.”11 Trantino, 752 A.2d at 780 (emphasis added) (quoting Final Report at *21). Because Holmes had remained “infraction free” after his initial hearing, J.A. at 52, these findings suggest that he would have enjoyed strong parole prospects under the pre-1997 rules. This inference is corroborated, at least to some extent, by an improvised data set that Holmes cites in his pro se complaint. Drawing on a string cite in a New Jersey Supreme Court opinion, Holmes identifies a group of felons who were sentenced to death before the 1970s and then had their sentences reduced to life in prison when New Jersey abolished the death penalty. J.A. at 15 (citing Trantino v. N.J. State Parole Bd., 764 A.2d 940, 947 n.2 (N.J. 2001); see also State v. Funicello, 286 A.2d 55, 58–59 (N.J. 1972). Holmes points out that under a parole scheme where the Board could not consider prospective parolees’ criminal history, these serious offenders were able to earn parole after fewer than 20 years in prison. From here, Holmes—who has spent many more years in prison—argues that had the Board limited its consideration to only new developments since his last hearing, he too could have earned parole already. Thus, we understand Holmes to allege that had he received similar treatment, it is plausible that he would have won release at the 2012 hearing. Unwilling to accept this point, the Government names twenty-two “convicted murderers” who received life sentences before the 1997 Amendments, obtained parole afterwards, and 11 While the Commission’s report casts new infractions as the sole basis for a successive parole denial under the pre- 1997 rules, the Government counters that the Board sometimes refused release based on psychological assessments. But the cases the Government cites concern parole hearings held after the 1997 Amendments came into effect. Those authorities therefore shed little light on the Board’s practices before the Amendments. In any event, the Commission’s central finding—that the grounds for a successive parole denial were extremely limited under the old rules—remains unchallenged. 15 did so about as quickly as the death-row inmates.12 Answering Br. at 25. If these prisoners represent the broader population of New Jersey homicide offenders, their experience implies that the Amendments made little difference to the Board’s propensity to grant parole. But if the Government selected these offenders because they secured release relatively quickly, their experience tells us nothing about the Amendments’ implementation. And, although we invited the Government to clarify how it created this list, it demurred. This failure to contextualize the data leaves us no choice but to discount it. None of this is to say that Holmes has conclusively established the requisite risk. To the contrary, all of his evidence is susceptible to multiple interpretations. Maybe the Commission overstated the practical effect of the pre-1997 rules on the practice of the Parole Board generally, or the death-row data is incomplete or otherwise misleading. After all, we observed in Royster that “the standards of the 1979 and 1948 Acts are identical with respect to inmates convicted before 1979.” 775 F.2d at 535. Maybe in Holmes’s particular case, the Board assigned less weight to old information than the written statement’s structure and substance suggest. At this stage, however, our obligation is to draw all reasonable inferences in Holmes’s favor. See Daniel, 766 F.3d at 61–62. When we do, Holmes’s complaint and the documents it incorporates by reference tell a plausible story: The old rules protected prisoners from repeated parole denials based on their criminal history; the 1997 Amendments removed that protection; and the Board relied partly on Holmes’s history in refusing to release him. So, while Holmes’s attack on the all- information provision’s implementation may hit rough water in discovery, it finds shelter enough to survive the Government’s motion to dismiss and to undergo discovery, which the District Court may wish to sequence, see Fed. R. Civ. P. 26(d)(3), as to the practical effect of the pre-1997 rules, 12 According to the Government, this list draws on data from the New Jersey Department of Corrections website. Whether to take judicial notice of that data is not a question we need to decide today, because even if we did so, it would not change our conclusion. 16 the Board’s consideration of past conduct in Holmes’s case, and whether their consideration created a significant risk of prolonging his imprisonment relative to the old rules. 2. The Risk Assessment-Requirement Apart from installing the all-information provision, the 1997 Amendments also instruct the Board to prepare “an objective risk assessment” before every parole hearing, including successive hearings. N.J. Stat. Ann. § 30:4- 123.52(e). Each assessment must survey an inmate’s “educational and employment background” and “family and marital history,” among other factors. Id. What this means in practice is that assessments often place old information before the Board. Holmes therefore argues that retroactively applying the risk-assessment requirement contradicts the Clause. We need not tarry long over this theory. To sustain it, Holmes must show that either the risk-assessment requirement’s terms or its implementation “individually disadvantaged” him. Richardson, 423 F.3d at 294. He has done neither. There is no serious argument that the risk-assessment requirement’s text establishes the requisite risk. See Garner, 529 U.S. at 255. By design, the requirement directs the Board to produce a new assessment of an inmate’s psychological state before each successive parole hearing. See N.J. Stat. Ann. § 30:4-123.52(e). That the assessments draw on a variety of sources, including old information, is not decisive. A current assessment of an inmate’s recidivism risk has always been part of the parole process. See N.J. Stat. Ann. § 30:4-123.56(c) (1996); see also Greenholtz v. Inmates of Neb. Penal and Corr. Complex, 442 U.S. 1, 10 (1979) (defining parole as an “assessment of . . . what a man is and what he may become”) (internal quotation marks and citation omitted). It is conceivable, of course, that the Board implemented the risk-assessment requirement in a way that harmed Holmes. If an assessment recites old information, and if the Board fixates on that information, an inmate might have a plausible ex post facto claim. Here, however, the Board’s written statement says nothing about old information embedded in the risk assessments—indeed, it does not discuss the assessments 17 at all. See J.A. at 45–54. Without more, we cannot conclude that the risk-assessment requirement played any role, let alone a “significant” one, in the Board’s calculus. Garner, 529 U.S. at 255. Where does that leave us? Of the many theories Holmes floats, only one merits discovery. The risk-assessment requirement fails to support a viable claim. The same is true of the all-information provision’s terms. But that provision’s implementation is a different story. As the Board’s written statement and historical practices attest, the provision plausibly created a significant risk of prolonging Holmes’s imprisonment. Having shown how Holmes’s claim reaches safe harbor, all that remains is to locate where the Government loses its way. C. The Government’s Counterarguments In urging us to chart a different course, the Government advances two arguments.13 In the first, it invites us to classify the 1997 Amendments as procedural changes exempt from ex post facto scrutiny. In the second, the Government submits that but for the 1997 Amendments, the Board would still have refused to release Holmes. Neither proposal is compatible with our precedents. 1. The Substantive-Procedural Distinction The Government’s initial argument borrows from a line of New Jersey Appellate Division cases. Under the Appellate Division’s approach, retroactively enforcing a “substantive” rule offends the Clause, but doing the same with a “procedural” rule does not. See Trantino, 752 A.2d at 780–81. Because the 1997 Amendments expand the evidence available to the Board, the Appellate Division deems them a “procedural 13 The Government declines to defend the District Court’s rationale for dismissing Holmes’s ex post facto claim. In rejecting that claim, the Court noted that “the Parole Board’s consideration of factors that suggest recidivism was consistent with the goals of either version of the statute.” Holmes II, 2018 WL 6522922, at *6. But our jurisprudence hinges on whether a new rule creates a significant risk of extended imprisonment, not on whether it aligns with broad penological principles. 18 modification” that lies outside the Clause’s reach. Id. at 781. This logic led the Appellate Division to dismiss Holmes’s ex post facto challenge, and the Government implores us to adopt a similar approach here. But a test that formalistically distinguishes between substantive rules and procedural ones finds no foundation in controlling cases or the functional approach that animates them. A review of three leading cases reveals that a challenged rule’s constitutionality hinges on its effect, not its form. In Morales, the Supreme Court scrutinized a rule’s “effect on [a] prisoner’s actual term of confinement.” 514 U.S. at 512 (emphasis added). In Richardson, we treated a rule’s “practical effect” as the touchstone of our inquiry. 423 F.3d at 291. And in Garner, the Supreme Court addressed a rule it described as “procedural”—yet that label played no part in the Court’s reasoning or result. 529 U.S. at 251, 254. Not only do controlling cases eschew formalist approaches generally, they expressly reject the specific tack the Government takes here. More than a century ago, the Court resolved a pair of ex post facto cases by deciding whether the challenged rules assumed substantive or procedural form. See Kring v. Missouri, 107 U.S. 221, 224 (1883); Thompson v. Utah, 170 U.S. 343, 351–52 (1898). But the Court has long since overruled those cases and renounced their reasoning. See Collins v. Youngblood, 497 U.S. 37, 46 (1990) (“[B]y simply labeling a law ‘procedural,’ a legislature does not thereby immunize it from scrutiny under the Ex Post Facto Clause.”) (citation omitted)). Time and again, the Court has refused “to define the scope of the Clause along an axis distinguishing between laws involving ‘substantial protections’ and those that are merely ‘procedural.’” Carmell v. Texas, 529 U.S. 513, 539 (2000). Instead, the “controlling inquiry” is whether a law “creates a significant risk of prolonging [the plaintiff’s] incarceration.” Garner, 529 U.S. at 250–51.14 14 Pennsylvania Prison Society v. Cortes is not to the contrary. See 622 F.3d 215 (3d Cir. 2010). It is true that Cortes condoned the dismissal of a claim partly because the challenged rules were “procedural and thus not ex post facto laws.” Id. at 234. But Cortes went on to clarify that the word 19 We appreciate that comity counsels caution before we part ways with New Jersey’s Appellate Division. Yet few of the Appellate Division’s cases grapple thoroughly with the Ex Post Facto Clause. And, despite our best efforts, we see no way to reconcile the Appellate Division’s formalist analysis with the functional approach embodied in Morales, Richardson, and Garner. So we cannot affirm on this basis. 2. The Board’s Review of New Information The Government’s fallback position fares no better. Had the Board restricted its review to new information, the Government says, it would still have refused to release Holmes. On the surface, this position seems to turn only on the new information before the Board. On closer examination, though, the Government’s argument enacts a subtle but substantial shift in the governing standard. By framing the test in terms of what the Board “would have” done, the Government conflates a significant-risk with a but-for cause. Arg. Tr. at 29. That contradicts controlling precedents, the Clause’s purposes, and comity principles. See Morales, 514 U.S. at 509; Mickens-Thomas, 321 F.3d at 391–92. And when we set aside the causation test, and instead focus on risk alone, the Board’s review of new information fails to foreclose Holmes’s claim. a) Where the But-for Test Goes Awry Three considerations convince us that a “but-for” test has “no basis in federal ex post facto law.” Richardson, 423 F.3d at 292 n.5. First off, that test amounts to a more demanding standard than one that turns on risk. To establish but-for causation, a plaintiff must demonstrate that a new rule probably accounted for an adverse parole decision. To satisfy the risk-based standard, however, a plaintiff need only show a “significant” possibility that a new rule prompted the parole “procedure” serves as shorthand for rules that present a “remote risk of impact on a prisoner’s expected term of confinement.” Id. at 237 (quoting Morales, 514 U.S. at 508). To the extent the Government construes Cortes more broadly—as announcing an alternative to the risk-based standard—its reading is inconsistent with Morales, Richardson, and Garner. 20 denial. Morales, 514 U.S. at 509; Mickens-Thomas, 321 F.3d at 392; see also Richardson, 423 F.3d at 292 (requiring a “significant” effect, rather than a “determinative” one). Confusing risk with causation thus artificially and inappropriately raises the bar for ex post facto claims. While the distinction between a significant risk and a but-for cause may seem minor, it carries major consequences. In parole cases, the Ex Post Facto Clause’s core mission is to ensure defendants understand their “chances of receiving early release” so “they can plea bargain and strategize effectively.” Mickens-Thomas, 321 F.3d at 391–92. But the causation test decouples our ex post facto analysis from this purpose. Suppose, for example, that a rule cuts a prisoner’s chances of release from 45% to 5%. That change would not constitute a but-for cause of an adverse parole decision, but it could easily inform a defendant’s trial and plea bargaining strategy. By excluding this and similar scenarios from constitutional protection, the causation test threatens to disable—or at least diminish—one of the Clause’s central objectives. Embracing a causation test would also endanger states’ authority over “confinement and release.” Garner, 529 U.S. at 252. Under a causation standard, federal courts would have to predict not just whether a challenged rule might have been significant, but whether it did control the outcome of a highly individualized and discretionary state proceeding. And just as federal courts’ reasoning would disturb state parole authority, so too would their results. Whenever a court finds that a change caused an adverse parole decision, the ensuing judgment would amount to a declaration that the parole board should grant release on remand. In sum, the Ex Post Facto Clause takes risk, not causation, as its touchstone. With that critical point clarified, we can make short work of the Government’s argument that new information dominated the Board’s decision. b) Why the Board’s Review of New Information Does Not Defeat Holmes’s Claim If we translate the Government’s argument into the register of risk, it fails. In staking out its position, the Government seizes on two types of new information. First, the 21 Board received several risk assessments that rated Holmes as a “high risk” of committing future crimes.15 J.A. at 40. Second, during a 2012 interview, Holmes “minimize[ed]” his past offenses and declined to accept “full responsibility” for them, an attitude the Board viewed as militating against release. Id. at 50–52. Taken together, the Government suggests, this new information left Holmes with such low odds of release that the Board’s review of old information made no real difference. This argument runs aground on a fundamental principle that undergirds the risk-based standard. To measure Holmes’s chances of receiving release, we avoid “post hoc” speculation, Mickens-Thomas, 321 F.3d at 387, and instead anchor our analysis in the Board’s “policy statements” and “actual practices,” Garner, 529 U.S. at 256. The Government overlooks this crucial lesson. Contrary to the Government’s suggestion, none of the Board’s practices conclusively establish that new information controlled its thinking. As noted, the written statement makes no mention of the risk assessments the Government insists the Board relied on. See Mickens-Thomas, 321 F.3d at 387, 389 (disregarding a “post hoc defense” that “did not appear in the formal Board decisions”). And, although the statement does discuss Holmes’s 2012 interview, it portrays that interview as a small piece of a larger puzzle. Nor is the interview’s substance so damaging as to deprive Holmes of any real hope for release. However disappointing, Holmes’s reluctance to accept “full responsibility” for decades-old crimes bears only indirectly on his propensity to commit future offenses. J.A. at 52; see N.J. Stat. Ann. § 30:4-123.56(c) (1996). That leaves open the possibility that other new information might have overcome the 2012 interview and convinced the Board to grant relief. To take one example, the Board’s statement highlights Holmes’s flawless disciplinary record over many years in prison. So had the Board restricted its review to that new information, a reasonable inference—though not the only one—is that Holmes’s perfect prison record could have outweighed his 15 For the sake of argument, we assume that the assessments constitute “new information” under the pre-1997 rule. N.J. Stat. Ann. § 30:4-123.56(c) (1996). 22 imperfect interview responses. To hold otherwise would disregard our pleading-stage duty to draw inferences in Holmes’s favor, not the Government’s. In the final analysis, then, neither counterargument scuppers Holmes’s ex post facto challenge. The formalistic substantive-procedural distinction collides with controlling cases. And when we analyze the new information before the Board under the rubric of risk, rather than causation, it fails to foreclose Holmes’s claim. Central to our conclusion, however, is that this case comes before us on the pleadings. With more information about the Board’s decision here, and about its practices more generally, the District Court may well reach a different result on remand. Our journey today shows that Holmes’s claim is seaworthy—not unsinkable. III. Due Process Clause To supplement his ex post facto claim, Holmes briefly invokes the Due Process Clause. On his account, the Clause mandates that parole decisions must be grounded in whatever rules governed at the time a prisoner committed a crime. But the Ex Post Facto Clause, not the Due Process Clause, establishes the relevant framework for resolving challenges to the retroactive application of new criminal rules. Neither the Due Process Clause’s procedural component nor its substantive one announces an anti- retroactivity principle. When a state creates a liberty interest in parole, as the Government concedes New Jersey did here,16 it cannot deprive prisoners of that interest without providing certain procedural protections. See Greenholtz, 442 U.S at 13, 16. The problem is that none of those protections prohibits the retroactive application of new rules. See id. As for the Clause’s substantive component, it controls only when a parole board considers a factor that “shocks the conscience.” Newman, 617 F.3d at 782 (internal quotation marks and citation omitted). And as our cases confirm, an inmate’s criminal history represents a common component of parole decisions, not a shocking or constitutionally-suspect one. See 16 We express no view as to the merits of the Government’s concession. 23 Block v. Potter, 631 F.2d 233, 238 (3d Cir. 1980); Greenholtz, 442 U.S. at 15. Extending the Due Process Clause to this domain also poses serious practical problems. Should Holmes’s theory prevail, the Clause would bar the retroactive application of any new parole rule. This promises to transform the Due Process Clause into an end-run around the Ex Post Facto Clause, which bars only the retroactive application of rules that reflect a “significant risk of increased punishment.” Richardson, 423 F.3d at 290. And because its roots lie in the Due Process Clause, this sweeping anti-retroactivity principle might extend not just to criminal cases, but also to civil ones. These profound consequences may explain why we have not found— and Holmes has not identified—any case invalidating the retroactive application of a new rule on due process grounds. The due process claim therefore finds no support in precedent or pragmatism, and we affirm its dismissal. IV. Conclusion For the foregoing reasons, we will vacate the dismissal of Holmes’s ex post facto claim, affirm the dismissal of his due process claim, and, as noted supra, remand for discovery to determine whether the retroactive application of the 1997 Amendments to Holmes “create[d] a significant risk of prolonging [his] incarceration.” Garner, 529 U.S. at 251. 24