concurring:
Our en banc decision in McKinney v. Ryan, 813 F.3d 798 (9th Cir. 2015) (en banc) (McKinney II), erred in concluding that' any Eddings error had a “substantial and injurious effect,” id, at 822 (quoting Brecht v. Abrahamson, 507 U.S. 619, 623, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993)), on the Arizona Supreme Court’s decision to affirm the defendant’s death sentence. State v. McKinney, 185 Ariz. 567, 917 P.2d 1214 (1996) (McKinney I). Asa result, our decision today is wrongly decided. Nevertheless, as a three-judge panel, wé are bound by McKinney II until either the Supreme Court or a future en banc panel overrules it. Therefore, I concur in the majority opinion and write separately only to point out how McKinney II’s error in applying Brecht infects our decision here.
I
Under AEDPA, we must determine whether the decision of the Arizona Supreme Court is contrary to or an unreasonable application of clearly established Supreme Court precedent. 28 U.S.p. § 2254(d)(1). It is clearly established that a sentencer may not “refuse to consider, as a matter of law, any relevant mitigating evidence.” Eddings v. Okla., 455 U.S. 104, 114, 102 S.Ct. 869, 71 L.Ed.2d 1 (1982) (italics in original); see also Lockett v. Ohio, 438 U.S. 586, 604, 98 S.Ct. 2954, 57 L.Ed.2d 973 (1978). While the sentencer “may determine the weight to be given relevant mitigating evidence,” it “may not give it no weight by excluding such evidence from [its] consideration.” Eddings, 455 U.S. at 114-15, 102 S.Ct. 869. Applying Lockett and Eddings, the Supreme Court held that a state cannot adopt a “causal nexus” rule,- that is, a rule precluding a sentencer from considering mitigating evidence unless there is a causal nexus between that evidence and the crime. Tennard v. Dretke, 542 U.S. 274, 287, 124 S.Ct. 2562, 159 L.Ed.2d 384 (2004). The sentencer may, however, consider “causal nexus ... as a factor in determining the weight or significance of mitigating evidence.” Lopez v. Ryan, 630 F.3d 1198, 1204 (9th Cir. 2011) overruled on other grounds by McKinney II, 813 F.3d at 819.
In this case, the Arizona Supreme Court stated only that it accorded no mitigating weight to Poyson’s evidence of mental health and an abusive childhood. State v. Poyson, 198 Ariz. 70, 81-82, 7 P.3d 79 (2000). Before McKinney II, we held that this decision was not an unreasonable application of Lockett, Eddings, and Tennard because we could not presume that the Arizona Supreme Court had refused to consider the mental health and abusive childhood evidence as a matter of law. See Poyson v. Ryan, 711 F.3d 1087, 1090 (9th Cir. 2013). Rather, as instructed by the Supreme Court, we adopted the “presumption that state courts know and follow the law.” Woodford v. Visciotti, 537 U.S. 19, 24, 123 S.Ct. 357, 154 L.Ed.2d 279 (2002); see Poyson, 711 F.3d at 1099.
*898McKinney II flipped this presumption. It held that we must presume the Arizona Supreme Court applied the unconstitutional causal nexus test between 1989 and 2005, even when, as here, the court expressly discussed the weight of the evidence. 813 F.3d at 803, 809, 816. This reasoning is contrary to Visciotti, as the McKinney II dissent made clear. See McKinney II, 813 F.3d at 827-860 (Bea, J,, dissenting). No further elaboration, of this error is needed.
II
I write separately to highlight McKinney II’s second error: its conclusion that a causal nexus, error has a “substantial and injurious effect” on a state court’s decision. 813 F.3d at 822-23.
A
Under Brecht, even if a state court unreasonably errs in applying Supreme Court precedent, a federal court may not provide habeas relief unless the error had a “substantial and injurious effect.” 507 U.S. at 623, 113 S.Ct. 1710. “There must be more than a ‘reasonable possibility* that the error was harmful.” Davis v. Ayala, — U.S. —, 135 S.Ct. 2187, 2198, 192 L.Ed.2d 323 (2015) (quoting Brecht, 507 U.S. at 637, 113 S.Ct. 1710). Rather, a “court must find that , the defendant was actually prejudiced by the error.” Id. (quoting Calderon v. Coleman, 525 U.S. 141, 146, 119 S.Ct. 500, 142 L.Ed.2d 521, (1998) (per curiam)). Even an Eddings error may be harmless., Greenway v. Ryan, 866 F.3d 1094, 1100 (9th Cir. 2017) (per curiam).
In determining that the Arizona Supreme Court’s presumed causal nexus error in McKinney I was prejudicial, McKinney II failed to provide a reasoned or reasonable -application of Brecht. Instead, without any meaningful analysis, McKinney II conclusorily held that the evidence presumed excluded under Arizona’s presumed causal nexus test “would have had a substantial impact on a capital sentencer who was permitted to evaluate and give appropriate weight to it as a nonstatutory mitigating factor.” McKinney II, 813 F.3d at 823. Therefore, McKinney II held, the Arizona Supreme Court’s “application of the test had a ‘substantial and injurious effect or influence’ on its decision to sentence [the defendant] to death.” Id. at 823-24 (quoting Brecht, 507 U.S. at 623, 113 S.Ct. 1710). In reaching this conclusion, McKinney II came close to enunciating a per se rule that when a state court’s application of a causal nexus test excludes mitigating evidence, such an error will not be harmless.
Such a quasi per se rule may be plausible when the sentencer in a particular case is a jury. If a state rule excludes certain mitigating evidence from the jury’s consideration as a matter of law, either the evidence will not be presented to the jury or the jury will be instructed to disregard it if they find no causal nexus. Because we presume a jury follows its instructions, Penry v. Johnson, 532 U.S. 782, 799, 121 S.Ct. 1910, 150 L.Ed.2d 9 (2001), and a jury generally does not give reasons for its decision, it is reasonable to presume that the jury could not meaningfully consider even strong mitigating evidence in reaching its verdict if it were excluded under a causal nexus rule, see Abdul-Kabir v. Quarterman, 550 U.S. 233, 255, 127 S.Ct. 1654, 167 L.Ed.2d 585 (2007); A court could determine that strong mitigating evidence which was excluded from consideration “would have had a substantial impact on a capital sentencer who was permitted to evaluate and give appropriate weight to it as a nonstatutory mitigating factor.” McKinney II, 813 F.3d at 823. Accordingly, in the absence of other factors (such as the presence of aggravating factors, that *899“overwhelmingly outweighed” the mitigating evidence, see Greenway, 866 F.3d at 1100), an Eddings error could have a substantial and injurious effect.
But the quasi per se rule adopted by McKinney II is entirely implausible when the sentencer is a state supreme court. Unlike a jury, a state supreme court has the authority to review and consider all the evidence in the record; this is particularly important, when as in Arizona, the state supreme court “reviews capital sentences de novo, making its own determination of what constitute legally relevant aggravating and mitigating factors, and then weighing those factors .independently.” McKinney II, 813 F.3d at 819 (citing Ariz. Rev. Stat. Ann. § 13-755). A state supreme court’s decision that certain categories of evidence are not mitigating is effectively the court’s conclusion that such evidence does not merit much weight. Just like a jury, a state supreme court can reasonably conclude that if a defendant’s mental impairments did not play a part in causing the defendant to commit a brutal offense, the impairments do not mitigate the defendant’s behavior.
A state supreme court’s conclusion about the mitigating weight of various types of evidence does not have the effect of excluding evidence as a matter of law. Nor does such a conclusion preclude a state supreme court from weighing the evidence differently in a different case. While a jury must follow instructions, the state court is free to disregard its instructions to itself because a state supreme court may always revisit its precedent. As the Arizona Supreme Court has explained, “while we should and do pay appropriate homage to precedent, we also realize that we are not prisoners of the past.” Lowing v. Allstate Ins. Co., 176 Ariz. 101, 107, 859 P.2d 724 (1993) (quoting Wiley v. The Indus. Comm’n of Ariz., 174 Ariz. 94, 103, 847 P.2d 595 (1993)). Indeed, even McKinney II acknowledged that by the mid-2000s, the Arizona Supreme Court had stopped applying the precedent that McKinney II presumed compelled the use of a causal nexus test, 813 F.3d at 817.
Finally, unlike a jury, a state supreme court generally' explains its reasons, and so may articulate its conclusion that defendant’s impairments merited little or no mitigating weight. See Greenway, 866 F.3d at 1100. Where a state supreme court has reached a reasoned conclusion that aggravating circumstances outweigh mitigating evidence in a particular case, there does not seem to be a reasonable possibility that the state supreme court would'reach a different result merely because a federal court announces that the state court has secretly maintained an unconstitutional causal nexus rule all along. See id.
B
Because McKinney II failed.to distinguish between a state supreme court and a jury, its Brecht analysis fails.
In McKinney I, the Arizona Supreme Court explained that it “conducts a thorough and independent review of the. record and of the aggravating and mitigating evidence to determine whether the sentence is justified, ... considering] the quality and strength, not simply the number, of aggravating or mitigating factors.” 185 Ariz at 578, 917 P.2d 1214. In its opinion, the Arizona Supreme Court reviewed the defendant’s evidence of childhood abuse and post-traumatic stress disorder (PTSD). Id. at 587, 917 P.2d 1214, 1226. It determined that the judge had fully considered evidence from several witnesses that defendant had “endured a terrible childhood,” as well as the PTSD diagnosis. Id. But the court held that, “a difficult family background, including childhood abuse, does not necessarily have substantial mitigating weight absent a showing that it significantly affected or impacted *900the defendant’s ability to perceive, comprehend, or control his actions.” Id. After considering the defendant’s abusive childhood and its impact on his behavior and ability to conform his conduct, the Arizona Supreme Court found there was no error in determining that the evidence of childhood abuse was “insufficiently mitigating to call for leniency.” Id.
In light of the Arizona Supreme Court’s reasoned consideration and weighing of the mitigating evidence, there was no basis for concluding that this same evidence would have a different impact—let alone a substantial impact—on the same court on resentencing simply because a federal court provides a reminder that Eddings precludes a sentencer from applying the causal nexus rule. McKinney II, 813 F.3d at 823-24. Brecht does not permit “mere speculation” about the potential prejudice to a defendant. Davis, 135 S.Ct. at 2198 (quoting Calderon, 525 U.S. at 146, 119 S.Ct. 500). Because there is not a reasonable possibility that the presumed legal error influenced the Arizona Supreme Court, or have more than a slight effect, the sentence should stand. See Kotteakos v. United States, 328 U.S. 750, 764, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946); Davis, 135 S.Ct. at 2198. McKinney II erred in ruling otherwise.
Ill
Because we are bound by McKinney IV s erroneous application of Brecht, its error infects this appeal as well. In our case, the Arizona Supreme Court considered Poyson’s mitigating evidence regarding his mental health and abusive childhood, but stated merely that it accorded these factors “no mitigating weight.” Poyson, 198 Ariz. at 81-82, 7 P.3d 79. On the other hand, the Arizona Supreme Court found that the evidence supported aggravating circumstances of (1) pecuniary gain, (2) especially cruel, heinous, or depraved murder, and (3) multiple homicide. Id. at 78-79, 7 P.3d 79. Based on its findings, the court upheld Poyson’s death sentence. Id. at 82, 7 P.3d 79. The court did so while performing its duty to “independently review and reweigh the aggravating and mitigating circumstances in every capital case ....’’Id. at 81, 7 P.3d 79.
Here, the Arizona Supreme Court reviewed and considered Poyson’s mitigating evidence, and balanced it against the case’s aggravating circumstances. Accordingly, there is no basis for concluding that our correction of any presumed Eddings error “would have had a substantial impact on a capital sentencer who was permitted to evaluate and give appropriate weight to it as a nonstatutory mitigating factor.” McKinney II, 813 F.3d at 823. We should therefore conclude that any presumed causal nexus error was not prejudicial, and therefore Poyson is not entitled to relief.
Because we are bound by McKinney II (at least for the time being), we are unable to reach this correct conclusion. As a result, I reluctantly concur in the majority opinion.