PER CURIAM Opinion; Concurrence by Judge KOZINSKI.
*1054OPINION
PER CURIAM.We consider whether the new rule announced in Richardson v. United States, 526 U.S. 813, 119 S.Ct. 1707, 143 L.Ed.2d 985 (1999), is retroactively applicable to criminal convictions that had become final by the time of the Supreme Court’s decision.
I
Montalvo was charged with one count of conspiracy to possess with intent to distribute cocaine, 21 U.S.C. § 846, and one count of engaging in a continuing criminal enterprise (CCE), 21 U.S.C. § 848. The jury convicted him of the CCE count, and the trial judge sentenced him to life imprisonment. Conspiracy being a lesser included offense of a CCE violation, Rutledge v. United States, 517 U.S. 292, 300, 116 S.Ct. 1241, 134 L.Ed.2d 419 (1996), the jury did not render a verdict on the conspiracy charge.
After numerous unsuccessful appeals on direct review, Montalvo’s conviction became final in 1996 when the Supreme Court denied his petition for certiorari. See Griffith v. Kentucky, 479 U.S. 314, 321 n. 6, 107 S.Ct. 708, 93 L.Ed.2d 649 (1987). Three years later, the Supreme Court decided Richardson v. United States, 526 U.S. 813, 119 S.Ct. 1707, 143 L.Ed.2d 985 (1999), where it interpreted, for the first time, the phrase “a continuing series of violations” in 21 U.S.C. § 848(c)(2).
The CCE statute reads:
[A] person is engaged in a continuing criminal enterprise if—
(1) he violates any provision of [these subchapters] the punishment for which is a felony, and
(2) such violation is a part of a continuing series of violations of [these subchapters]—
(A) which are undertaken by such person in concert with five or more other persons with respect to whom such person occupies a position of organizer, a supervisory position, or any other position of management, and
(B) from which such person obtains substantial income or resources.
21 U.S.C. § 848(c) (emphasis added).
Richardson held that a jury “must unanimously agree not only that the defendant committed some ‘continuing series of violations’ but also that the defendant committed each of the individual ‘violations’ necessary to make up that ‘continuing series.’ ” 526 U.S. at 815, 119 S.Ct. 1707 (emphasis added). If a “series” requires three or more violations, United States v. Valenzuela, 596 F.2d 1361, 1367 (9th Cir.1979), Richardson demands that the jury unanimously agree on at least three specific predicate felonies that constitute the “continuing series” before it may convict.
Here, the government alleged that Montalvo was involved in thirteen felony drug offenses. Suppose Jurors I-VI found that Montalvo had committed crimes 1-3 and nothing else, and Jurors VII-XII found that Montalvo had committed crimes 4-6 and nothing else. Under Richardson’s unanimity requirement, the jury could not return a CCE verdict — even if all the jurors agree that Montalvo, in one way or another, committed at least three drug violations. Only if twelve jurors all agree on at least three specific predicate offenses (say, offenses 1, 4 and 5) may they convict under the CCE statute. Furthermore, the court must give a “specific unanimity” instruction to this effect. Richardson, 526 U.S. at 816, 119 S.Ct. 1707.
Montalvo’s trial took place before Richardson, and the trial judge did not give the *1055instruction that Richardson would subsequently require.1 Soon after Richardson was decided and while his habeas petition was still pending before the district court, Montalvo filed a “partial reply” to the government’s opposition to his petition and challenged the jury instruction as inadequate. The district court denied the writ, holding that Montalvo’s Richardson claim could not be asserted under the anti-retroactivity principle of Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989).
II
“[N]ew constitutional rules of criminal procedure will not be applicable to those cases which have become final before the new rules are announced, unless the new rule [1] places certain kinds of primary, private individual conduct beyond the power of the criminal law-making authority to proscribe, or [2] could be considered a watershed rule of criminal procedure.” Bousley v. United States, 523 U.S. 614, 619-20, 118 S.Ct. 1604, 140 L.Ed.2d 828 (1998) (internal quotation marks and citations omitted); see Teague v. Lane, 489 U.S. 288, 311-12, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989) (plurality opinion). In short, Teague’s anti-retroactivity principle “applies only to procedural rules”; “it is inapplicable to the situation in which [the Supreme Court] decides the meaning of a criminal statute enacted by Congress.” Bousley, 523 U.S. at 620, 118 S.Ct. 1604. The dispositive question here, then, is whether Richardson’s requirement — which admittedly is a “new” rule under Teague, 489 U.S. at 301, 109 S.Ct. 1060—is substantive or procedural.2
Every one of our sister circuits that have confronted the issue agrees that Richardson announced a new substantive rule of criminal law. See Santana-Madera v. United States, 260 F.3d 133, 138-39 (2d Cir.2001); United States v. Lopez, 248 F.3d 427, 431-32 (5th Cir.2001); Murr v. United States, 200 F.3d 895, 904-06 (6th Cir.2000); Lanier v. United States, 220 F.3d 833, 838 (7th Cir.2000); United States v. Barajas-Diaz, 313 F.3d 1242, 1245 (10th Cir.2002); Ross v. United States, 289 F.3d 677, 681 (11th Cir.2002) (per curiam).
We join that unanimous view. In Richardson, the Supreme Court was faced with the question whether the phrase “series of violations” refers to a single element—namely, a “series”—with the individual “violations” merely constituting the underlying brute facts or “means” as to which jury unanimity is not required, see Andersen v. United States, 170 U.S. 481, 499-501, 18 S.Ct. 689, 42 L.Ed. 1116 (1898), or whether the phrase “series of violations ” creates several elements—namely, the discrete “violations”—with respect to each of which the jury must agree unanimously and separately, see Andres v. United States, 333 *1056U.S. 740, 748, 68 S.Ct. 880, 92 L.Ed. 1055 (1948). Richardson v. United States, 526 U.S. at 817-24, 119 S.Ct. 1707. After examining the statute’s language in the context of what Congress was trying to accomplish, the Supreme Court deemed the latter more consistent with our tradition of jury unanimity. Id.
What’s significant for our purposes is that, in analyzing what constitutes “elements” as opposed to brute facts or “means,” Richardson was “decid[ing] the meaning of a criminal statute enacted by Congress,” Bousley, 523 U.S. at 620, 118 S.Ct. 1604. Richardson is, in this respect, analytically indistinguishable from Bailey v. United States, 516 U.S. 137, 116 S.Ct. 501, 133 L.Ed.2d 472 (1995), where the Supreme Court interpreted the meaning of the word “use” in a criminal statute that imposes a five-year minimum sentence upon any person who “during and in relation to any crime of violence or drug trafficking crime ... uses or carries a firearm,” 18 U.S.C. § 924(c)(1)(A). Bailey similarly looked to “language, context, and history” before concluding that the element “use” in the statute requires “active employment of the firearm by the defendant, a use that makes the firearm an operative factor in relation to the predicate offense.” Bailey, 516 U.S. at 143-44, 116 S.Ct. 501. Just as Bailey narrows the reach of the firearms statute, Richardson restricts the class of persons who can be convicted for CCE—i.e., only those individuals who committed at least three predicate drug felonies, as determined by a unanimous jury beyond a reasonable doubt. Because the new rule announced in Bailey is substantive and therefore retroactively applicable to criminal convictions that have become final, Bousley, 523 U.S. at 619-21, 118 S.Ct. 1604, we reach the same conclusion about the rule established by Richardson. See also United States v. Mitchell, 867 F.2d 1232, 1233 (9th Cir.1989) (per curiam) (holding that McNally v. United States, 483 U.S. 350, 356-60, 107 S.Ct. 2875, 97 L.Ed.2d 292 (1987), which narrowed the federal mail fraud statute to protect only property rights, is retroactive); United States v. Sood, 969 F.2d 774, 775-76 (9th Cir.1992) (holding that United States v. Bordado, 857 F.2d 519, 524 (9th Cir.1988), which narrowed the federal bribery statute to exclude application to Guam, is retroactive); United States v. McClelland, 941 F.2d 999, 1001 (9th Cir.1991) (holding that United States v. Aguon, 851 F.2d 1158 (9th Cir.1988) (en banc), which held “inducement” to be an element of the federal extortion statute, is retroactive).
We hold that Teague does not bar Montalvo from raising a Richardson claim in his section 2255 habeas petition.
III
We must next decide whether the trial judge’s failure to give a Richardson specific unanimity instruction was harmless.3
A. We can quickly reject Montalvo’s argument that Richardson violations are structural errors subject to the automatic reversal rule. The “general rule” is that constitutional errors do not require reversal of a conviction, but are *1057susceptible to harmless error review. Arizona v. Fulminante, 499 U.S. 279, 306-07, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991) (collecting cases). The list of structural defects is short and limited: “total deprivation of the right to counsel at trial,” see Gideon v. Wainwright, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963); “a judge who was not impartial,” see Tumey v. Ohio, 273 U.S. 510, 47 S.Ct. 437, 71 L.Ed. 749 (1927); “unlawful exclusion of members of the defendant’s race from a grand jury,” see Vasquez v. Hillery, 474 U.S. 254, 106 S.Ct. 617, 88 L.Ed.2d 598 (1986); “the right to self-representation at trial,” see McKaskle v. Wiggins, 465 U.S. 168, 104 S.Ct. 944, 79 L.Ed.2d 122 (1984); “the right to [a] public trial,” see Waller v. Georgia, 467 U.S. 39, 104 S.Ct. 2210, 81 L.Ed.2d 31 (1984). Fulminante, 499 U.S. at 309-10, 111 S.Ct. 1246; see also Sullivan v. Louisiana, 508 U.S. 275, 113 S.Ct. 2078, 124 L.Ed.2d 182 (1993) (holding that a constitutionally deficient reasonable doubt instruction cannot be harmless error).
Because Richardson errors affect only the way the jury is instructed on an element of the offense, they do not taint the trial “from beginning to end” or undermine “the framework within which the trial proceeds.” Fulminante, 499 U.S. at 309-10, 111 S.Ct. 1246. The Supreme Court has held that “an instruction that omits an element of the offense” altogether “does not necessarily render a criminal trial fundamentally unfair,” and the error is therefore susceptible to harmless-error analysis. Neder v. United States, 527 U.S. 1, 9, 119 S.Ct. 1827, 144 L.Ed.2d 35 (1999). A fortiori, an instruction that merely omits the unanimity requirement with respect to an element of the offense cannot be categorized as structural. This conclusion is consistent with the unanimous view of our sister circuits. See Ross v. United States, 289 F.3d 677, 681-82 (11th Cir.2002); Santana-Madera v. United States, 260 F.3d 133, 139 (2d Cir.2001); United States v. Jefferson, 215 F.3d 820, 823 (8th Cir.2000); United States v. Escobar-de Jesús, 187 F.3d 148, 161-62 (1st Cir.1999).
There remains the question of which harmless error review standard applies: Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967) (“[B]efore a federal constitutional error can be held harmless, the court must be able to declare a belief that it was harmless beyond a reasonable doubt.”), or the more deferential Brecht v. Abrahamson, 507 U.S. 619, 623, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993) (requiring “substantial and injurious effect or influence in determining the jury’s verdict” before the conviction may be set aside) (quoting Kotteakos v. United States, 328 U.S. 750, 776, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946))?
We held in Bains v. Cambra, 204 F.3d 964 (9th Cir.2000), that the more deferential Brecht standard “should apply uniformly in all federal habeas corpus cases under § 2254.” Id. at 977. Montalvo argues that the reasons behind Bains’s adoption of Brecht in section 2254 cases are inapplicable here. In Montalvo’s view, the “interests laid out in Brecht,” such as “finality of state court convictions, comity, federalism, and prominence of the trial itself ... are notably lacking in federal habeas review of federal convictions under 28 U.S.C. § 2255.” Appellant’s Opening Br. at 17 n. 14 (internal quotation marks omitted).
While Montalvo is right that federalism interests are not implicated in the section 2255 context, we do not read Brecht’s “substantial and injurious effect” standard as resting primarily on federalism concerns. It is, more than anything else, motivated by “considerations underlying our habeas jurisprudence,” Brecht, 507 U.S. at 633, 113 S.Ct. 1710, that apply equally to collateral attacks on both feder*1058al and state convictions—considerations such as the different purposes of direct and collateral review and the importance of finality in criminal convictions. Our reading of Brecht is fortified by the case’s citation to Teague’s anti-retroactivity principle to illustrate the general practice of “appl[ying] different standards on habeas than [those] on direct review.” Id. at 634, 113 S.Ct. 1710. We have already held that Teague applies equally to sections 2254 and 2255 habeas cases. United States v. Sanchez-Cervantes, 282 F.3d 664, 667-68 (9th Cir.2002). We hold now that Brecht’s harmless error standard applies to habeas cases under section 2255, just as it does to those under section 2254. See also Murr v. United States, 200 F.3d 895, 906 (6th Cir.2000) (applying Brecht on federal habeas review of defendant’s CCE conviction).
B. The district court denied Montalvo’s petition only because it held — contrary to our conclusion in Part II, supra— that Montalvo’s Richardson claim was Teague-barred. It further held, however, that if Teague did not bar Montalvo from asserting the claim on collateral review, the error was not harmless. After reviewing the record de novo, see Ghent v. Woodford, 279 F.3d 1121, 1126 (9th Cir.2002), we conclude that the trial court’s failure to instruct the jury on Richardson’s specific unanimity requirement did not have a “substantial and injurious effect” on the jury’s verdict.
The district court started down the right track when it noted that there are at least two circumstances where a Richardson jury instruction error is per se harmless: first, if the jury that convicted the defendant of a CCE charge also convicted him of at least three related drug violations alleged to be predicate violations constituting the “continuing series”; second, if the district court instructed the jury that a “continuing series of violations” means “three or more violations,” and the government introduced evidence of only three predicate offenses at trial. In both of these scenarios, the jury would have necessarily agreed unanimously that defendant committed at least three specific predicate felonies that make up the “continuing series.” See United States v. Brown, 202 F.3d 691, 700 (4th Cir.2000).
The district court veered off course when it held that, because neither of these scenarios is present here—i.e., there was no other jury verdict beyond Montalvo’s CCE conviction, and the government presented not only three, but thirteen, drug violations—“[i]t is unknowable from the record of this complex case which acts the jury unanimously agreed upon, [and] it is possible that some jurors found some acts to have been committed, while other jurors found others, but the jury was able to agree unanimously only that there were three [or more] that were committed.” Order at 43 (emphasis added).
That all twelve jurors did not necessarily find Montalvo guilty of three specific predicate drug felonies is not dispositive. See California v. Roy, 519 U.S. 2, 4, 117 S.Ct. 337, 136 L.Ed.2d 266 (1996) (per curiam) (holding that a harmless error analysis that asks whether “the jury necessarily found the omitted element” is “far too strict” under Brecht). The question is “whether, in light of the record as a whole,” the error had a “substantial and injurious effect or influence in determining the jury’s verdict.” Brecht, 507 U.S. at 638, 113 S.Ct. 1710 (emphasis added and internal quotation marks omitted). This inquiry is not a matter of pure logic, but of sensitive judgment based on the case presented to the jury by the prosecution and the defense.
Here, the government’s case against Montalvo was overwhelming. In addition to witness testimony, there was documen*1059tary evidence showing Montalvo’s involvement in at least thirteen instances of drug distribution. See, e.g., Gov’t Exhibits 601A-C. The thirteen discrete offenses were never contested individually, but were treated as a group. In fact, Montalvo didn’t even dispute the existence of the whole conspiracy, only the extent of his own involvement. As Montalvo’s counsel summed up the defense’s theory, there was a conspiracy and a series of drug distributions alright, but everything was masterminded by the Woodhouse brothers who wanted to pin blame on Montalvo in exchange for reduced sentences: “Montalvo did not understand the intricacies. He did not understand the Woodhouse operation. The fact that he was at the Woodhouse properties, properties that he arranged for Woodhouse, the fact that he may even had some suspicion does not make him guilty of anything.” Reporters Transcript at 2298; see also id. at 2283 (“My suggestion to you, and I believe the evidence supports that this is actually a Woodhouse operation.”); id. at 2287 (“Obviously, if you believe Woodhouse, what else is there to talk about?”); id. at 2297 (“Greg Woodhouse was the accountant who paid the money, gave instructions, set up routes, made a decision as to which people would do what in terms of customers, not Montalvo.”). Furthermore, according to the defense, the Woodhouses’ attempt to blame Montalvo was helped by an ATF agent’s vendetta against the defendant.
Because we have already rejected Montalvo’s various other challenges to his conviction on direct appeal, see Montalvo I, 972 F.2d 1346, 1992 WL 184342 (9th Cir. Aug.4, 1992); Montalvo II, No. 92-16532 (9th Cir. Nov. 30, 1992); Montalvo III, 74 F.3d 1247, 1996 WL 2205 (9th Cir. Jan. 3, 1996); Montalvo IV, 172 F.3d 60, 1999 WL 51547 (9th Cir. Jan. 27, 1999), the guilty verdict necessarily means that all twelve jurors found that Montalvo committed at least three predicate felonies (though there is a logical possibility that they did not agree on which ones), that he supervised five or more persons and that he obtained substantial income or resources. See 21 U.S.C. § 848(c). Montalvo’s defense at trial focused exclusively on undermining the government’s case with respect to the position he occupied in the conspiracy’s hierarchy — a defense the jury rejected— not the thirteen discrete drug transactions. The Richardson error was therefore harmless. See Brown, 202 F.3d at 700-01 (holding that a Richardson error is harmless if “the element was uncontested and supported by overwhelming evidence” (footnote omitted)). We are not free to disturb the jury’s credibility determinations and its finding that Montalvo was the boss.
IV
In sum, we affirm the district court’s denial of Montalvo’s petition, but on different grounds. The Richardson claim was not Teague-barred and could be asserted on collateral review. Nevertheless, the error did not have a “substantial and injurious effect” on the jury’s verdict under Brecht.
AFFIRMED.
. Although the trial court did not instruct the jury that it must reach unanimity with respect to at least three specific drug felonies that make up the "continuing series," it did give a general unanimity instruction as follows:
It is very important for you to attempt to reach a unanimous verdict, but, of course, only if each of you can do so after having been — if after having made your own conscientious independent decision. Do not change an honest belief about the weight and effect of the evidence simply to reach a verdict.
Reporter’s Transcript at 2404.
. Because we hold that Richardson fits within Teague’s first exception, we need not address the applicability of the second exception—whether Richardson counts as a "watershed rule[ ] of criminal procedure,” Teague, 489 U.S. at 311, 109 S.Ct. 1060, that "alter[s] our understanding of the bedrock procedural elements [essential to] the fairness of a particular conviction." Mackey v. United States, 401 U.S. 667, 693, 91 S.Ct. 1160, 28 L.Ed.2d 404 (1971) (Harlan, J., concurring in part, dissenting in part).
. We held in United States v. Hernandez-Escarsega, 886 F.2d 1560, 1572-73 (9th Cir.1989), that failure to instruct the jury on the specific unanimity requirement could be harmless. We also held specifically in Montalvo’s direct appeal, Montalvo I, 972 F.2d 1346, 1992 WL 184342 (9th Cir. Aug.4, 1992), that he was not prejudiced by the jury instruction. These decisions, however, preceded Richardson. We now consider whether the analysis changes after Richardson. See United States v. Scrivner, 189 F.3d 825, 827 (9th Cir.1999) (holding that law of the case does not bar reconsideration when there has been an intervening change in the law).