dissenting:
In 1985, Hector Juan Ayala shot and killed three men. In 1989, he was convicted of three counts of murder, and the jury returned a death sentence. On direct appeal his conviction and sentence were affirmed by the California Supreme Court in 2000. People v. Ayala, 24 Cal.4th 243, 99 Cal.Rptr.2d 532, 6 P.3d 193 (2000). The Supreme Court of the United States denied his petition for certiorari in 2001. Ayala v. California, 532 U.S. 908, 121 S.Ct. 1235, 149 L.Ed.2d 143 (2001). Ayala filed his initial petition for a writ of habeas corpus in the United States District Court for the Southern District of California in 2002. This appeal is from the district court’s February 17, 2009, final order denying the petition.1
The majority holds, based primarily on law developed after Ayala’s trial, that Ayala must be released or retried because it suspects the prosecutor might have had a racial motive in recusing seven jurors. It does so by inappropriately deconstructing the California Supreme Court’s opinion to justify its evasion of the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”) so that it may review the state trial court’s 1989 decisions de novo. Because Ayala’s federal claim is Teague-barred, and because the majority’s approach and conclusion are contrary to AEDPA and recent Supreme Court opinions, I dissent.
I
I agree with the district court and the State that Ayala’s claim that he was deprived of his constitutional rights when his attorney was not present when the prosecutor offered his reasons for the challenged recusals, is barred under Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989).2
A. The standard for determining whether a claim is barred under Teague.
In Caspari v. Bohlen, 510 U.S. 383, 114 S.Ct. 948, 127 L.Ed.2d 236 (1994), the Su*869preme Court set forth the test for determining whether a state prisoner’s claim was Teague-barred:
“[A] case announces a new rule if the result was not dictated by precedent existing at the time the defendant’s conviction became final.” Teague v. Lane, supra, 489 U.S. at 301 [109 S.Ct. 1060], In determining whether a state prisoner is entitled to habeas relief, a federal court should apply Teague by proceeding in three steps. First, the court must ascertain the date on which the defendant’s conviction and sentence became final for Teague purposes. Second, the court must “[s]urve[y] the legal landscape as it then existed,” Graham v. Collins, supra, 506 U.S. at 468 [113 S.Ct. 892], and “determine whether a state court considering [the defendant’s] claim at the time his conviction became final would have felt compelled by existing precedent to conclude that the rule [he] seeks was required by the Constitution,” Saffle v. Parks, 494 U.S. 484, 488 [110 S.Ct. 1257, 108 L.Ed.2d 415] (1990). Finally, even if the court determines that the defendant seeks the benefit of a new rule, the court must decide whether that rule falls within one of the two narrow exceptions to the nonretroactivity principle. See Gilmore v. Taylor, 508 U.S. 333, 345 [113 S.Ct. 2112, 124 L.Ed.2d 306] (1993).
Id. at 390, 114 S.Ct. 948 (emphasis as quoted in Caspari).3
There is no dispute that Ayala’s conviction became final in May 2001, when the Supreme Court denied certiorari, and Ayala does not assert that he comes within either of the two narrow exceptions. The remaining question is whether, in May 2001, the unconstitutionality of ex parte procedure used by the trial court in 1986 was “dictated” by precedent.
B. The evolution of Batson as of May 2001.
In Batson v. Kentucky, 476 U.S. 79, 96, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986), the Supreme Court held that “a defendant may establish a prima facie case of purposeful discrimination in selection of the petit jury solely on evidence concerning the prosecutor’s exercise of peremptory challenges at the defendant’s trial.” See also Georgia v. McCollum, 505 U.S. 42, 47, 112 S.Ct. 2348, 120 L.Ed.2d 33 (1992). Batson established a three-step inquiry. First, the defendant must make a prima facie showing that the prosecution has exercised peremptory challenges in a racially discriminatory manner. Batson, 476 U.S. at 96, 106 S.Ct. 1712. The Supreme Court stated that it had “confidence that trial judges, experienced in supervising voir dire, will be able to decide if the circumstances concerning the prosecutor’s use of peremptory challenges creates a prima facie case of discrimination against black jurors.” Id. at 97, 106 S.Ct. 1712. Second, “[o]nce the defendant makes a prima facie showing, the burden shifts to the State to come forward with a neutral explanation for challenging black jurors.” Id. Third, the trial court must then “determine if the defendant has established purposeful discrimination.” Id. at 98, 106 S.Ct. 1712.
In setting forth this three-step standard, the Supreme Court specifically declined “to formulate particular procedures to be followed upon a defendant’s timely objection to a prosecutor’s challenges.” Id. at 99, 106 S.Ct. 1712. The Court reiterated that “[i]n light of the variety of jury selection practices followed in our state and federal trial courts, we make no attempt to instruct these courts how best to implement our holding today.” Id. at 99 n. 24, 106 S.Ct. 1712. As a result, during the *870quarter of a century that has passed since Batson, courts have considered numerous ways of applying Batson’s three-step standard.
Ayala’s primary argument is that the trial court’s exclusion of him and his counsel from the proceedings in which the prosecution justified its recusal of seven jurors violated his constitutional rights to assistance of counsel at critical stages of the proceedings, to be personally present, and to assist his counsel in his defense. In response, the State argued and the district court held that in 2001, when Ayala’s conviction became final, the exclusion of Ayala and his counsel from the proceedings was not a constitutional violation, and hence, Ayala’s claim was barred by Teague, 489 U.S. 288, 109 S.Ct. 1060.
The California Supreme Court, in reviewing Ayala’s direct appeal, concluded that it was “almost universally recognized that ex parte proceedings following a motion regarding peremptory challenges allegedly made on the basis of improper group bias are poor procedure and should not be conducted unless compelling reasons justify them.” Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 203.
The majority claims that in May 2001 this rule had been “unequivocally ‘dictated by precedent’ ” as a result of our opinion in United States v. Thompson, 827 F.2d 1254 (9th Cir.1987). Majority at pp. 869-70. Thompson concerned a 1985 criminal trial in a federal district court. The judge alone conducted voir dire and “the government used four of its peremptory challenges to exclude all four blacks in the venire.” Id. at 1256. When Thompson’s lawyer moved for a mistrial, the district court “allowed the government to put its reasons for the disputed peremptory challenges on the record, albeit in camera and out of the presence of the defendant and his lawyer.” Id. Thompson appealed, arguing that this procedure violated his Fifth Amendment right to due process and his Sixth Amendment right to a fair and impartial jury. Id. A divided panel concluded that the “district court erred in refusing to allow defense counsel in this case to hear the government’s reasons for excluding the black potential jurors and to present argument thereon.” Id. at 1261. We explained that “situations where the court acts with the benefit of only one side’s presentation are uneasy compromises with some overriding necessity, such as the need to act quickly or to keep sensitive information from the opposing party. Absent such compelling justification, ex parte proceedings are anathema in our system of justice and, in the context of a criminal trial, may amount to a denial of due process.”4 Id. at 1258-59.
C. Ayala’s Sixth Amendment claim was not dictated by precedent in 2001.
In Horn v. Banks, 536 U.S. 266, 272, 122 S.Ct. 2147, 153 L.Ed.2d 301 (2002), the Supreme Court held that “a federal court considering a habeas petition must conduct a threshold Teague analysis when the issue is properly raised by the state,” even if the state supreme court did not consider the issue. Thus, we are required to determine as a threshold matter whether Ayala’s claim is Teague-barred. I would hold that Ayala’s claim is Teague-barred because it was not “dictated” by Supreme Court case law, and the case Ayala relies upon, Thompson, did not announce a clear constitutional rule. The majority confuses the wisdom of Thompson with whether that wisdom had been embraced by 2001.
*871Thompson is not a Supreme Court opinion and concerned a federal court trial, not a state court trial. Accordingly, it could not dictate the result in Ayala’s case when his conviction became final. In Massachusetts Delivery Ass’n v. Coakley, 671 F.3d 33, 47 (1st Cir.2012), the First Circuit reiterated that “[s]tate courts are not bound by the dictates of the lower federal courts, although they are free to rely on the opinions of such courts when adjudicating federal claims.” (internal citations omitted.). Similarly, in Bromley v. Crisp, 561 F.2d 1351, 1354 (10th Cir.1977), the Tenth Circuit noted that “the Oklahoma Courts may express their differing views on the retro-activity problem or similar federal questions until we are all guided by a binding decision of the Supreme Court.” Also, in U.S. ex rel. Lawrence v. Woods, 432 F.2d 1072, 1076 (7th Cir.1970), the Seventh Circuit agreed that “because lower federal courts exercise no appellate jurisdiction over state tribunals, decisions of lower federal courts are not conclusive on state courts.” Consistent with our sister circuits’ decisions, our prior opinion in Thompson, which was an appeal from a federal criminal action, was not binding on the California Supreme Court.
Furthermore, even though the logic behind the opinion in Thompson may be compelling, the opinion does not set forth a clear rule of constitutional law. The opinion recognized that there were “occasional departures from” the norm of holding adversarial proceedings, noted a number of instances in which in camera proceedings were appropriate, and concluded that departure from the norm “may amount to a denial of due process.” Id. at 1258-59 (emphasis added). The language in Thompson is vague compared, for example, to our statement in Menefield v. Borg, 881 F.2d 696, 699 (9th Cir.1989), that “we hold that the right to counsel attaches to the motion for a new trial stage.”
The fact that Thompson did not lay down a clear rule of constitutional law is confirmed by a review of other Ninth Circuit cases as well as decisions by our sister circuits. In Lewis v. Lewis, 321 F.3d 824, 831 n. 27 (9th Cir.2003), we observed that “[cjertainly, requiring a court to allow defense counsel to argue [during the three-step Batson process] is not clearly established law.” Thus, fifteen years after Thompson issued, we did not think that Thompson established a clear rule of constitutional law. In Majid v. Portwondo, 428 F.3d 112 (2d Cir.2005), the Second Circuit commented that “[i]t remains at least arguable that courts holding Batson hearings may, to the contrary, hear the explanations in camera and outside the presence of the defendants.” Id. at 128 (citations omitted). In United States v. Tucker, 836 F.2d 334, 340 (7th Cir.1988), the Seventh Circuit noted that the Supreme Court in Batson expressly declined to formulate procedures and disagreed with the Ninth Circuit’s opinion in Thompson insofar as it required “adversarial hearings once a defendant establishes a prima facie case of purposeful discrimination.” Similarly, in United States v. Davis, 809 F.2d 1194, 1202 (6th Cir.1987), the Sixth Circuit commented that Batson does not “require the participation of defense counsel while the Government’s explanations are being proffered.”
The majority strives mightily to distinguish these cases on the grounds that they are not well-reasoned, in some instances are dicta, and have been rejected by other circuits and most state courts. But the standard established by the Supreme Court for determining whether an issue is Teague-barred is not the merits of the old rule, or even recognition of the wisdom of the new rule, but whether the new rule was “dictated by precedent.” Caspari, 510 U.S. at 390, 114 S.Ct. 948. These conflict*872ing cases confirm that Thompson did not dictate the result in Ayala’s case.
Moreover, as noted, the rule that the majority claims was established in Thompson is not a bright-line rule. Rather, at most, Thompson states that defense counsel could not be excluded absent some “compelling justification.” See Majority at p. 866. Here, the prosecutor offered an explanation for seeking to present his reasons in camera: he did not want to reveal his strategy to the defense. Following Thompson and the other cases cited by the majority, it is now clear that this is not a valid reason not to follow the norm of an adversarial proceeding.5 However, this was not firmly established in 2001 when Ayala’s conviction became final.
In sum, I agree with the district court that the right to be present and have counsel present when the prosecution presented its reasons for its challenged recu-sals was not “dictated by precedent” when Ayala’s conviction became final, and therefore the issue is Teague-barred.
II
Assuming the issue is not Teague-barred, we must next turn to the question of what exactly the California Supreme Court held and what deference it is owed. The court first acknowledged that “no particular procedures are constitutionally required” to conduct a Batson hearing, thereby rejecting Ayala’s federal claim. Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 202. “The next question,” it said, was “whether it was error to exclude defendant from participating in the hearings on his Wheeler motions.” It concluded that “as a matter of state law, it was.” Id. 99 Cal.Rptr.2d 532, 6 P.3d at 203 (emphasis added). After conducting a thorough review of relevant state and federal precedents, it reiterated that it had “concluded that error had occurred under state law” and “noted” Thompson’s suggestion that the error may amount to a denial of due process. Id. 99 Cal.Rptr.2d 532, 6 P.3d at 204. Nonetheless, it concluded that the error was “harmless under state law, and that, if federal error occurred, it, too, was harmless beyond a reasonable doubt as a matter of federal law.” Id. (internal citations omitted). Thus, although there may be some question as to whether the California Supreme Court actually found that there was federal error, it clearly addressed Ayala’s federal claim in determining that whatever federal error occurred, it was harmless as a matter of federal law.
The majority and I part ways as to how to review this holding. Because the state court adjudicated Ayala’s federal claim on the merits and rejected it, we must accord that decision deference under AEDPA. See Johnson v. Williams, — U.S. -, 133 S.Ct. 1088, 1094, 185 L.Ed.2d 105 (2013) (“[Wjhen a federal claim has been presented to a state court and the state court has denied relief, it may be presumed that the state court adjudicated the claim on the merits in the absence of any indication or state-law procedural principles to the contrary, (internal citations omitted)”).
However, the majority’s dislike for AEDPA drives it to try to avoid its provi*873sions. In its initial opinion, the majority interpreted the Supreme Court’s opinion in Brecht v. Abrahamson, 507 U.S. 619, 113 5.Ct. 1710, 123 L.Ed.2d 353 (1993), in such a manner as to allow it to grant relief without deferring to the California Supreme Court. See Ayala v. Wong, 693 F.3d 945, 961-63 (9th Cir.2012). The majority now takes a different tack in an effort to circumnavigate AEDPA and review Ayala’s 1989 state court conviction de novo. Contrary to the Supreme Court’s recent opinions, the majority deconstructs the California Supreme Court’s opinion. In doing so, the majority: (1) separates the California Supreme Court’s finding of error under state law from its adjudication of the federal claim; (2) decides that the California Supreme Court did not determine whether there was error under federal law; and then (3) concludes that it has “no reason to give § 2254(d) deference” to the California Supreme Court’s decision. Majority at p. 845.
The majority’s approach is fundamentally flawed for at least two reasons. First, it is contrary to the Supreme Court’s opinions directing that any question as to whether a state court considered a constitutional issue is to be resolved in favor of finding that it did. Harrington v. Richter, — U.S.-, 131 S.Ct. 770, 784-85, 178 L.Ed.2d 624 (2011); Johnson v. Williams, — U.S.-, 133 S.Ct. 1088, 1094-96, 185 L.Ed.2d 105 (2013). Second, by separating the California Supreme Court’s determination of Batson/Wheeler6 error from its adjudication of the federal claim, the majority evades giving the opinion the deference demanded by AEDPA and the United States Supreme Court.
A. Supreme Court precedent compels the conclusion that the California Supreme Court decided Ayala’s federal claims on their merits.
1. The California Supreme Court’s opinion.
The California Supreme Court’s evaluation of the Batson/Wheeler issue was clear and concise. It held that “it was error to exclude defendant from participating in the hearings on the Wheeler motions.” Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 203. Indeed, the California Supreme Court recognized that the error enhanced “the risk that defendant’s inability to rebut the prosecution’s stated reasons will leave the record incomplete.” Id. 99 Cal.Rptr.2d 532, 6 P.3d at 204. As the majority admits, the California Supreme Court cited both Batson and our opinion in Thompson in concluding that “it seems to be almost universally recognized that ex parte proceedings following a motion regarding peremptory challenges allegedly made on the basis of improper group bias are poor procedure and should not be conducted unless compelling reasons justify them.” Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 203.
The California Supreme Court then turned to the question of prejudice and held:
We have concluded that error occurred under state law, and we have noted Thompson’s suggestion that excluding the defense from a Wheeler-type hearing may amount to a denial of due process. We nonetheless conclude that the error was harmless under state law (People v. Watson (1956) 46 Cal.2d 818, 836, 299 P.2d 243), and that, if federal error occurred, it, too, was harmless beyond a reasonable doubt (Chapman v. California (1967) 386 U.S. 18, 24 [87 *874S.Ct. 824, 17 L.Ed.2d 705]) as a matter of federal law.
99 Cal.Rptr.2d 532, 6 P.3d at 204. There can be no doubt that the California Supreme Court adjudicated the federal claim, first by implicitly rejecting it, and then, as an alternative holding, finding that any error was harmless.
2. The majority’s deconstruction of the California Supreme Court’s opinion.
The majority acknowledges this portion of the California Supreme Court’s opinion. Majority at pp. 844-46. However, it then proceeds to mull over whether the state court (a) held there was error under federal constitutional law, (b) held there was no error under federal constitutional law, or (3) did not decide whether there was error under' federal constitutional law. Id. These are idle musings, for the “only question that matters under § 2254(d)(1)” is whether the petitioner’s claim was adjudicated on the merits, and whether that adjudication was contrary to or an unreasonable application of clearly established Supreme Court precedent. Lockyer v. Andrade, 538 U.S. 63, 71, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003). Whether one agrees with the California Supreme Court’s decision or not, the federal claim was clearly adjudicated.
The majority, although purporting to accept that the California Supreme Court found constitutional error, proceeds to argue that “alternatively” the exception to deference set forth in Richter applies.7 Majority at pp. 842-43. Exactly how the majority reaches this conclusion is not clear. Here is its explanation:
Richter and Williams instruct us to afford a rebuttable presumption that a fairly presented claim was “adjudicated on the merits” for purposes of § 2254(d), but this presumption is rebuttable if there is “any indication or state-law procedural principles” supporting the conclusion that the state court did not adjudicate the federal claim on the merits. Richter, 131 S.Ct. at 784-85. Here, the California Supreme Court denied Ayala relief overall but did so by (1) finding that the trial court committed error on state law grounds, (2) failing to make any express determination of error on federal constitutional grounds, and (3) finding any error harmless under both the state and federal standards for harmless error. In the context of these holdings, the rebuttable presumption that Richter and Williams instruct us to afford is, in fact, rebutted. The California Supreme Court, by finding any alleged error harmless under both the state and federal standards for harmless error, had no reason to reach the question of whether federal constitutional error occurred.
Majority at p. 842.
The majority then argues that the California Supreme Court “would have had good reason not to decide the merits of the *875federal constitutional issue” because courts generally do not pass on questions of constitutionality unless adjudication is unavoidable. Majority at p. 20. Also, by analogy it draws on cases involving claims of ineffective assistance of counsel under Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), where the Supreme Court has interpreted a state court’s silence as a failure to reach an issue and thus not an adjudication on the merits. Majority at pp. 843-44.
The majority then proffers its vision of the law: “There are circumstances in which, even if a state court has denied relief overall, a state court’s silence with respect to a fairly presented federal claim cannot be interpreted as an ‘adjudication on the merits’ for purposes of § 2254(d), because the rebuttable presumption cited in Richter and Williams is rebutted by the legal principles involved (including the principle of constitutional avoidance) and factual context applicable to a particular case.” Majority at p. 844. This novel approach allows the majority to assert that the “California Supreme Court had no reason to reach Ayala’s federal constitutional claim,” and to conclude that it has “no reason to give § 2254 deference” to the California Supreme Court’s decision. Majority at p. 873.
3. The majority’s approach is contrary to recent Supreme Court opinions.
The majority’s deconstruction of the California Supreme Court’s opinion, besides being unnecessary dicta and unpersuasive, is contrary to recent Supreme Court opinions that were directed at the Ninth Circuit. In both Richter, 131 S.Ct. 770, and Johnson, 133 S.Ct. 1088, the Supreme Court reversed us for failing to give proper deference to the state courts’ opinions.8 Moreover, the Supreme Court set forth a clear standard that leaves no doubt that here the California Supreme Court considered Ayala’s federal claims on their merits.
In Richter, the Court stressed that AEDPA “bars relitigation of any claim ‘adjudicated on the merits in state court,’ subject only to the exceptions in §§ 2254(d)(1) and (d)(2).” 131 S.Ct. at 784 (internal quotation marks omitted). The Court noted that there does not need to be “an opinion from the state court explaining the state court’s reasoning,” and the state court need not say it was adjudicating the federal claim on its merits. Id. It further held that, “[wjhen a federal claim has been presented to a state court and the state court has denied relief, it may be presumed that the state court adjudicated the claim on the merits in the absence of any indication or state-law procedural principles to the contrary.” Id. at 784-85.
Two years later, in Williams, a unanimous Supreme Court found it necessary to remind us of this standard. Williams challenged his conviction for first degree murder on the ground that the trial court improperly dismissed a juror. 133 S.Ct. at 1093. The California state courts denied him relief and the district court denied Williams’ habeas petition. Judge Reinhardt, writing for a three judge panel of the Ninth Circuit, held that despite the Richter presumption, the state courts had not adjudicated Williams’ federal claim,9 *876applied a de novo standard of review, and granted relief. Williams, 646 F.3d at 641, 653.
The Supreme Court firmly rejected our opinion. It first noted that the assumption that a federal claim was overlooked by the state court is wrong for a number of reasons, including: (1) “there are circumstances in which a line of state precedent is viewed as fully incorporating a related federal constitutional right,”10 Williams, 133 S.Ct. at 1094; (2) “a state court may not regard a fleeting reference to a provision of the Federal Constitution or federal precedent as sufficient to raise a separate federal claim,” id. at 1095; and (3) “there are instances in which a state court may simply regard a claim as too insubstantial to merit discussion.” Id. The Supreme Court concluded that, “[w]hen a state court rejects a federal claim without expressly addressing that claim, a federal habeas court must presume that the federal claim was adjudicated on the merits.” Id. at 1096.
If federal courts are to presume that a state court considers a federal claim even when the court does not expressly address the claim, it follows that where, as here, the California Supreme Court’s opinion clearly reflects that the court was aware of the federal claims, we must accept that the federal claims were “adjudicated on the merits,” and limit any relief according to § 2254(d). See Richter, 131 S.Ct. at 784.
The majority, however, seizes on the Supreme Court’s statement that the presumption “can in some limited eircum-stances be rebutted,” Williams, 133 S.Ct. at 1096, and adopts a definition of the exception that swallows the presumption. The majority asserts that the presumption does not apply because “it was not necessary for the state court to reject the federal constitutional claim on the merits in order for it to deny relief to the petitioner,” and because “the facts of this case dictate the conclusion that the California Supreme Court believed that the error under state law also constituted federal constitutional error.” Majority at p. 841. The first reason is contrary to Williams because the Supreme Court held that the presumption applies even where the state court regards a federal claim “as too insubstantial to merit discussion.” 133 S.Ct. at 1095. The second proffered reason is also contrary to Williams because the Court held that the presumption applies where “state precedent is viewed as fully incorporating a related federal constitutional right.” Id. at 1094.
Furthermore, in order to rebut the Richter presumption there must be “reason to think some other explanation for the state court’s decision is more likely.” Richter, 131 S.Ct. at 785. But here, there can be no doubt that the California Supreme Court did consider Ayala’s federal claims, first by stating that “no particular procedures are constitutionally required” to hold a Batson hearing, and then, in the alternative, holding that any federal error was harmless. Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 202-04. Thus, the California Supreme Court’s opinion is not best read *877as addressing Ayala’s federal claims, as the majority admits,11 but can only be so read.
In sum, a review of the California Supreme Court’s opinion allows for only one conclusion: that the court considered Ayala’s federal claims. Moreover, even if this conclusion was not mandated, and the presumption set forth by the Supreme Court in Richter and Williams came into play, the presumption would be controlling. Because the California Supreme Court considered Ayala’s federal claims (and, in any event, must be presumed to have done so), the AEDPA standard of review applies.
B. The majority applies a de novo standard of review and fails to give proper deference to the California Supreme Court’s opinion.
The majority’s treatment of the California Supreme Court’s ruling on the Bat-sonfWheeler violation is a smoke screen designed to obscure the fact that the majority reviews prejudice de novo rather than under the AEDPA deference standard. This approach cannot be squared with the Supreme Court’s recent opinions, which require that we ask whether the California Supreme Court’s determination that the error was harmless beyond a reasonable doubt was “so lacking in justification that there was an error well understood and comprehended in existing law beyond any possibility of fair minded disagreement.” Richter, 131 S.Ct. at 786-87.
1. Deference is mandated by the statute and Supreme Court precedent.
The applicable provisions of AEDPA are codified in 28 U.S.C. § 2254.12 The Supreme Court’s opinions hold that the AEDPA standard applies to the California Supreme Court’s finding of harmless error. In Richter, the Supreme Court reiterated that federal habeas relief is not available “unless it is shown that the earlier state court’s decision ‘was contrary to’ federal law then clearly established in the holdings of this Court,” or “ ‘was based on an unreasonable determination of the facts’ in light of the record before the state court.” 131 S.Ct. at 785 (internal citations omitted). Thus, it does not matter whether the California Supreme Court’s determination of harmless error is a question of fact, or law, or mixed; it is entitled to deference under AEDPA.
The Supreme Court further elaborated on the applicable standard. In Richter, it *878held that a “state court’s determination that a claim lacks merit precludes federal habeas relief so long as ‘fairminded jurists could disagree’ on the correctness of the state court’s decision.” 131 S.Ct. at 786 (citing Yarborough v. Alvarado, 541 U.S. 652, 664, 124 S.Ct. 2140, 158 L.Ed.2d 938 (2004)). In addressing prejudice in a claim of ineffective assistance of counsel under Strickland, the Supreme Court held:
[A] challenger must demonstrate “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different. A reasonable probability is a probability sufficient to undermine confidence in the outcome.” [Strickland v. Washington, 466 U.S. 668] at 694 [104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) ]. It is not enough “to show that the errors had some conceivable effect on the outcome of the proceeding.” Id. at 693 [104 S.Ct. 2052]. Counsel’s errors must be “so serious as to deprive the defendant of a fair trial, a trial whose result is reliable.” Id. at 687 [104 S.Ct. 2052],
131 S.Ct. at 787-88. The Supreme Court again emphasized this deference when it reversed us in Felkner v. Jackson, — U.S.-, 131 S.Ct. 1305, 179 L.Ed.2d 374 (2011).13
2. The majority’s alternate standard of review is contrary to Supreme Court precedent.
The majority, however, invokes Brecht, 507 U.S. 619, 113 S.Ct. 1710, in order to justify what is, in essence, a de novo standard of review. Majority at pp. 849-51. The majority asserts that relief may be granted if we find that the error had a “substantial and injurious effect or influence in determining the jury’s verdict,” Majority at p. 849 (quoting Brecht, 507 U.S. at 623, 113 S.Ct. 1710), or “if one cannot say, with fair assurance, after pondering all that happened without stripping the erroneous action for the whole, that the judgment was not substantially swayed by the error,” Majority at p. 850 (quoting Merolillo v. Yates, 663 F.3d 444, 454 (9th Cir.2011)), or even where a “judge feels himself in virtual equipoise as to the harmlessness of the error.” Majority at p. 850 (quoting Merolillo, 663 F.3d at 454). That the majority essentially conceives of this as de novo review is obvious from its declaration that “[i]f we cannot say that the exclusion of defense counsel with or without the loss of the questionnaires likely did not prevent Ayala from prevailing on his Bat-son claim, then we must grant the writ.” Majority at p. 851.
Brecht was decided before the passage of AEDPA. In Fry v. Pliler, 551 U.S. 112, *879121, 127 S.Ct. 2321, 168 L.Ed.2d 16 (2007), the Supreme Court held that “in § 2254 proceedings a court must assess the prejudicial impact of constitutional error in a state-court criminal trial under the ‘substantial and injurious effect’ standard set forth in Brecht.” However, in doing so, the Supreme Court construed the Brecht standard as including the AEDPA standard:
Three years after we decided Brecht, Congress passed, and the President signed, the [AEDPA], under which a habeas petition may not be granted unless the state court’s adjudication “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.... ” 28 U.S.C. § 2254(d)(1). In Mitchell v. Esparza, 540 U.S. 12 [124 S.Ct. 7, 157 L.Ed.2d 263] (2003) (per curiam), we held that, when a state court determines that a constitutional violation is harmless, a federal court may not award habeas relief under § 2254 unless the harmlessness determination itself was unreasonable. Petitioner contends that § 2254(d)(1), as interpreted in Esparza, eliminates the requirement that a petitioner also satisfy Brecht’s standard. We think not. That conclusion is not suggested by Esparza, which had no reason to decide the point. Nor is it suggested by the text of AEDPA, which sets forth a precondition to the grant of habeas relief (“a writ of habeas corpus ... shall not be granted” unless the conditions of § 2254(d) are met), not an entitlement to it. Given our frequent recognition that AEDPA limited rather than expanded the availability of habeas relief, see, e.g., Williams v. Taylor, 529 U.S. 362, 412 [120 S.Ct. 1495, 146 L.Ed.2d 389] (2000), it is implausible that, without saying so, AEDPA replaced the Brecht standard of “ ‘actual prejudice,’ ” 507 U.S. at 637 [113 S.Ct. 1710] (quoting United States v. Lane, 474 U.S. 438, 449 [106 S.Ct. 725, 88 L.Ed.2d 814] (1986)), with the more liberal AEBPAJChapman standard which requires only that the state court’s harmless-beyond-a-reasonable-doubt determination be unreasonable. That said, it certainly makes no sense to require formal application of both tests (AED-TA/Chapman and Brecht) when the latter obviously subsumes the former.
551 U.S. at 119-20, 127 S.Ct. 2321.
Three aspects of the Supreme Court’s explanation are particularly important. First, the Court endorsed its prior opinion in Esparza, 540 U.S. 12, 124 S.Ct. 7, that habeas relief was available only if the state court’s determination of harmlessness was unreasonable. Fry, 551 U.S. at 119, 127 S.Ct. 2321. Second, the Court reiterated that AEDPA “limited rather than expanded the availability of habeas relief.” Id. Third, the Court held that the Brecht “actual prejudice” standard requires a greater showing than the “the more liberal AED-PA/Chapman standard which requires only that the state court’s harmless-beyond-a-reasonable-doubt determination be unreasonable.” Id. at 119-20, 127 S.Ct. 2321. See also Brecht, 507 U.S. at 637, 113 S.Ct. 1710 (holding that habeas petitioners “are not entitled to habeas relief based on trial error unless they can establish that it resulted in ‘actual prejudice’ ”); Pulido v. Chrones, 629 F.3d 1007, 1020 (9th Cir. 2010) (holding that because the petitioner “did not suffer any actual prejudice, he is not entitled to habeas relief’).
The majority attempts to evade the deference inherent in the AEDPA¡Brecht standard by quoting language from Mero-lillo, 663 F.3d at 454. The majority asserts that when a judge is “in virtual equipoise as to the harmlessness of the error” and has “grave doubt about whether an *880error affected a- jury [substantially and injuriously], the judge must treat the error as if it did so.” Majority at p. 850 (citations omitted).
The majority takes this standard out of context. Merolillo does not suggest that the state court’s opinion is not entitled to deference. Our opinion first recognized that we continue to look to the last reasoned decision of the state court, and that the state court’s findings “are entitled to a presumption of correctness unless the petitioner rebuts the presumption with clear and convincing evidence.” 663 F.3d at 453 (citing 28 U.S.C. § 2254(e)(1)). Although the opinion refers to the Supreme Court’s prior elucidations on harmless error, we concluded that Brecht’s “substantial and injurious effect” standard governs our harmless error review. Merolillo, 663 F.3d at 455. Our opinion also determined that applying the AEDPA/Chapmcm standard, the state court’s determination of harmless error was objectively unreasonable. Id
3. Deference to the state court opinion reconciles Brecht and the Supreme Court’s recent opinions.
The majority fails to appreciate that even when focusing on harmlessness, a state court’s factual findings are entitled to deference, see, e.g., Mansfield v. Sec’y, Dep’t of Corr., 679 F.3d 1301, 1309 (11th Cir.2012), and that this deference informs the definition of “grave doubt.” In order to grant relief, a federal court must have “grave doubt” as to the harmlessness of the error. Critically, the Supreme Court’s opinions mandate that this “grave doubt” be objective rather than subjective. Our personal perspectives as to harmlessness are not controlling. Rather, we are directed to consider whether “there is no possibility fairminded jurists could disagree” with the state court’s decision, and whether there “was an error well understood and comprehended in existing law beyond any possibility for fairminded disagreement.” Richter, 131 S.Ct. at 786-87. Thus, by definition, where fairminded jurists can disagree, there is no “grave doubt” as to the harmlessness of the error. In other words, in state habeas petitions subject to AEDPA, any objective “grave doubt” as to the harmlessness of an error is dispelled by a determination that “fair-minded jurists” could disagree.
The reach of this mandate from Richter can be illustrated by considering the majority’s statement in a footnote. The majority opines that the California Supreme Court’s conclusion that Ayala was not prejudiced was “an unreasonable application of Chapman.14, Chapman, decided some 30 years before the passage of AEDPA, held “that before 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.” Id. at 24, 87 S.Ct. 824. AEDPA does not change the constitutional standard, but it does alter the standard applied by federal courts when reviewing a state prisoner’s habeas petition. AEDPA limits relief to instances where the state court adjudication was “an unreasonable application of, clearly established Federal law,” or “an unreasonable determination of the facts in light of the evidence.” 28 U.S.C. § 2254(d). The Supreme Court’s opinions in cases such as Richter and Williams provide an objective measure for what is unreasonable under *881AEDPA: a state court’s determination of a federal claim is unreasonable only if no fairminded jurist could agree with it.
In this case, consistent with the Supreme Court’s opinions, a writ may not issue just because “we cannot say that the exclusion of defense counsel and the loss of questionnaires likely did not prevent Ayala from prevailing on his Batson claim.” Majority at p. 850. Rather, a writ may issue only if we determine (using the majority’s language) that there is a “grave doubt as to the harmlessness of the error,” meaning that no fairminded jurist could find that the exclusion of defense counsel and the loss of questionnaires did not prevent Ayala from prevailing on his Batson claim.15 See, e.g., Richter, 131 S.Ct. at 786-87. As set forth in the next section, fairminded jurists can concur in the California Supreme Court’s determination of harmless error.
Ill
A review of the record shows that although the loss of the questionnaires and the exclusion of defense counsel constitute error, fairminded jurists can agree with the California Supreme Court that those facts did not prevent Ayala from prevailing on his Batson claim.
A. The loss of certain prospective jurors’ questionnaires.
The majority states that it is “unable to evaluate the legitimacy of some of the prosecution’s proffered reasons for striking the black and Hispanic jurors because they referred to questionnaires that are now lost.” Majority at p. 852. Of course, this statement misses the mark because the real question is whether the record was sufficient to allow the California Supreme Court to review Ayala’s claims as he presented them to that court. A review of the California Supreme Court’s opinion and Ayala’s filings shows that the state court fully and fairly considered his claims. Furthermore, Ayala has not shown that the loss of certain prospective jurors’ questionnaires violated his constitutional rights or that the loss prejudiced him.
First, it is critical to note what was in the record before the California Supreme Court. The record contained the voir dire transcript for all prospective jurors, the transcript of the in camera hearings on the prosecutor’s reasons for the recusals, the questionnaires of all the seated jurors, and the questionnaires of the alternate jurors. What was missing were the 77-question, 17-page questionnaires the 200 or so other potential jurors had filled out.
In Boyd v. Newland, 467 F.3d 1139 (9th Cir.2006), we recognized that the Supreme Comet’s opinion in Miller-El v. Dretke, 545 U.S. 231, 125 S.Ct. 2317, 162 L.Ed.2d 196 (2005), holds that “comparative juror analysis is an important tool that courts should utilize in assessing Batson claims.”16 *882Boyd, 467 F.3d at 1145. We commented that “comparative juror analysis” referred “to an examination of a prosecutor’s questions to prospective jurors and the jurors’ responses, to see whether the prosecutor treated otherwise similar jurors differently because of their membership in a particular group.” Id. Boyd concluded:
A reviewing court cannot examine the “totality of the relevant facts” and “all relevant circumstances,” Batson, 476 U.S. at 94 [106 S.Ct. 1712], surrounding a prosecutor’s peremptory strike of a minority potential juror without an entire voir dire transcript. A transcript of the complete voir dire, as distinct from a partial transcript up to the time of the Batson motion, is proper because comparative juror analysis is appropriate both at the time of the Batson motion and in light of all subsequent voir dire testimony.
467 F.3d at 1151. Here, we have the entire voir dire transcript. Moreover, there is nothing in Boyd to suggest that in addition to the voir dire transcript, juror questionnaires from jurors who are not selected are essential to a determination of the totality of the relevant facts.
Indeed, the opposite conclusion can be drawn from our treatment in Boyd of a California rule requiring an indigent defendant to show some cause in order to receive a free transcript of voir dire. We held, citing United States v. MacCollom, 426 U.S. 317, 322-23, 96 S.Ct. 2086, 48 L.Ed.2d 666 (1976), that the California rule did not violate the Constitution, but that the state court erred in failing to recognize that the defendant had raised a plausible Batson claim entitling him to a transcript of voir dire. Boyd, 467 F.3d at 1151. If a defendant can be required to show some cause in order to receive a transcript of voir dire, it follows that a defendant has no per se right to the preservation of all questionnaires filled out by prospective jurors who were not seated.
To be fair, there is language in our en banc opinion in Kesser v. Cambra, 465 F.3d 351 (9th Cir.2006) (en banc), that could be read to infer a right to juror questionnaires. We concluded that: “In this case, an evaluation of the voir dire transcript and juror questionnaires clearly and convincingly refutes each of the prosecutor’s nonracial grounds, compelling the conclusion that his actual and only reason for striking [a juror] was her race.” Id. at 360. But this statement shows only that where juror questionnaires are available, we will consider them, not that the questionnaires are necessary. Instead, we commented that a comparative juror analysis was appropriate because “[w]e too have a transcript of voir dire and a Batson claim fairly presented, and that is all Miller-El requires.” Id. at 361.
In addition, we recently commented on the lack of questionnaires of excused jurors in Briggs v. Grounds, 682 F.3d 1165 (9th Cir.2012). Briggs concerned a Batson challenge to a state court conviction where the federal record did not contain the questionnaires of excused jurors. Id. at 1170. In affirming the district court’s denial of relief, the majority noted:
The dissent seems to conclude that because we cannot independently verify the answers from the questionnaires as they are not in the record, the defense’s characterization is equally, if not more, plausible despite the state court determinations to the contrary. However, “AEDPA imposes a highly deferential standard for evaluating state-court rulings and demands that state-court decisions be given the benefit of the doubt,” Felkner v. Jackson, [sic, — U.S.-], 131 S.Ct. 1305, 1307 [179 L.Ed.2d 374] (2011) (per curiam) (internal quotation *883marks omitted) (overturning the Ninth Circuit). The dissent’s readiness to doubt the state court determination based on the defendant’s characterization of the record does not apply the appropriate level of deference Congress and the United States Supreme Court have required of us.
Id. at 1170-71. The majority in Briggs further noted that “it is widely acknowledged that the trial judge is in the best position to evaluate the credibility of the prosecutor’s proffered justifications.” Id. at 1171 (internal citations omitted). Citing the Supreme Court’s statements in Rice v. Collins, 546 U.S. 333, 338-39, 126 S.Ct. 969, 163 L.Ed.2d 824 (2006), that a “federal habeas court can only grant Collins’ petition if it was unreasonable to credit the prosecutor’s race-neutral explanations for the Batson challenge,” we stated in Briggs that:
it would be anathema to AEDPA if we were to assume that the petitioner’s contentions about the questionnaires are true simply because the record before us does not contain the excused jurors’ questionnaires. The burden to disprove the factual findings rests with Briggs. 28 U.S.C. § 2254(e)(1) (requiring “clear and convincing evidence” to rebut “a determination of a factual issue made by a State court”).
Id.
It follows that the lack of prospective jurors’ questionnaires does not relieve Ayala of his burden to show by clear and convincing evidence that the California Supreme Court was wrong in determining that the prosecutor was not biased. Accordingly, we must determine whether Ayala has shown that the lack of these questionnaires in his case renders the record insufficient for a determination of his federal claim. He fails in this task for several reasons.
First, the California Supreme Court reasonably rejected Ayala’s claim that his constitutional rights were infringed by the loss of the bulk of prospective juror questionnaires. It explained:
The deficiency of which he complains is the absence of certain questionnaires, which were completed by prospective jurors, then lodged with the superior court, subsequently lost by its clerk’s office, and finally determined by the superior court to be beyond reconstruction. A criminal defendant is indeed entitled to a record on appeal that is adequate to permit meaningful review. That is true under California law. [Citation.] It is true as well under the United States Constitution—under the Fourteenth Amendment generally, and under the Eighth Amendment specifically when a sentence of death is involved. [Citation.] The record on appeal is inadequate, however, only if the complained-of deficiency is prejudicial to the defendant’s ability to prosecute his appeal. ([People v. Alvarez, 14 Cal.4th 155,] at p. 196, fn. 8 [58 Cal.Rptr.2d 385, 926 P.2d 365 (1996) ]).
Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 208. The California Supreme Court concluded that if the loss of the questionnaires was error under either federal or state law, “it was harmless beyond a reasonable doubt.” Id. This determination is reasonable and entitled to deference. Rice, 546 U.S. at 338-39, 126 S.Ct. 969.
Second, the determination is reasonable because the missing juror questionnaires are not critical to Ayala’s federal claims. The questionnaires of the 70 or so jurors who-were never called are not relevant because Ayala does not allege, let alone show, that the potential jurors were excused due to constitutionally forbidden reasons.
*884Third, none of the prosecutor’s stated reasons for recusing the questioned jurors relied solely, or even primarily, on the jurors’ questionnaires. Rather, in each instance the prosecutor mentioned the juror’s specific answers to questions posed on voir dire. In a couple of instances the prosecutor referenced a person’s questionnaire, but this was primarily to explain why the prosecutor found the individual’s oral responses troubling.
Finally, Ayala has ably presented his specific Batson challenges based on the voir dire transcript and the extant questionnaires of the seated jurors and alternates. Although Ayala argues that the lost questionnaires might support his arguments, such a contention can be made about any lost document. If such speculation constituted prejudice, the standard would be reduced to a per se rule.
B. Challenges to the Individual Jurors
The remaining issue is whether Ayala has shown by clear and convincing evidence that no reasonable jurist could have credited the prosecutor’s non-diseriminato-ry reasons for excusing the seven jurors in issue. In other words, whether at least one fairminded jurist could agree with the California Supreme Court’s opinion. The majority discusses only three of the jurors in its opinion, but a review of the prosecutor’s reasons for excusing each of the seven jurors shows that the California Supreme Court’s determination that “the challenged jurors were excluded for proper, race-neutral reasons” was reasonable. See Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 204.
1. Olanders D.
Olanders D. was one of the first jurors challenged by Ayala. The trial court held that Ayala had not met the first prong of the Batson test (a prima facie showing that the challenge was based on race, see Kesser, 465 F.3d at 359), but nonetheless indicated that it would hear the prosecutor’s reasons for the recusal in order to have a complete record. The prosecutor stated in the ex parte proceeding:
My primary concern with regard to [Olanders D.] is his ability to vote for the death [sentence] during the penalty phrase. On his questionnaire he indicated that he does not believe in the death penalty. He did indicate that his view had changed over the last several years. He told us that he did want to serve. During the time that he was questioned, I felt that his responses were not totally responsive to the questions of either counsel for the defense or myself.
My observations in reading his questionnaire and before even making note of his racial orientation was that his responses on the questionnaire were poor. They were not thought out. He demonstrated a lack of ability to express himself well. And his answers did not make a lot of sense. As a result, I felt that he is not a person who could actively participate in a meaningful way in deliberations with other jurors, and his ability to fit in with a cohesive group of 12 people I sincerely question, and it was for that reason plus his stand on the death penalty that led me to believe that I did not want him on this jury.
The trial judge responded:
Okay. Certainly with reference to whether or not he would get along with 12 people, it may well be that he would get along very well with 12 people. I think the other observations of counsel are accurate and borne out by the record.
The California Supreme Court held that the record showed that the challenged jurors were excluded for proper, race-neutral causes. Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 204. Addressing Olanders D., the court commented:
*885[T]he prosecutor stated he had exercised the challenge in part because his questionnaire indicated he opposed the death penalty. The prosecutor acknowledged Olanders D.’s oral statements that his views had changed, but commented that his answers were “not totally responsive to the questions of either counsel for the defense or myself.” He further stated, in essence, that Olanders D.’s difficulties in communicating led him to question whether he would “fit in” on the jury. The court disagreed with the latter point, noting, “it may well be that he would get along very well with 12 people,” but added: “I think the other observations of counsel are accurate and borne out by the record.”
Id. The California Supreme Court further noted that the trial court “credited the prosecutor’s opinion[] that Olanders D. opposed the death penalty.” Id. 99 Cal.Rptr.2d 532, 6 P.3d at 206.
The majority claims that the prosecutor’s motives for excusing Olanders D. is suspect for several reasons. First, Ayala “could have pointed to seated white jurors” who similarly expressed hesitancy to impose the death penalty. Majority at p. 853. Second, the majority asserts that its review of the voir dire transcript shows that “Olanders D.’s answers were responsive and complete.” Majority at p. 854. Third, it claims that the responses of a seated white juror were just as unresponsive. Majority at p. 854. The majority concludes that none of the reasons proffered by the prosecutor should be sustained because one was rejected by the trial judge, “two failed to distinguish the juror whatsoever from at least one seated white juror,” and the fourth cannot be evaluated because his questionnaire was lost. Majority at pp. 854-55.
Were we reviewing the trial judge’s decision de novo, the majority’s approach might be persuasive. But the applicable standard is whether no fairminded judge could agree with the California Supreme Court’s determination that the juror was excluded for proper, race-neutral reasons. See Richter, 131 S.Ct. at 786. Ayala does not come close to meeting this standard.
There is no suggestion that any seated juror raised a similar set of concerns as Olanders D. The trial judge, who had the opportunity to observe Olanders D., agreed with the prosecutor that Olanders D. was ambivalent about the death penalty, had not been responsive on his questionnaire, and lacked the ability to express himself clearly. Moreover, the trial judge did not necessarily reject the prosecutor’s concern that Olanders D. could not participate in a meaningful way in jury deliberations, but rather only commented that it “may well be that he would get along very well with 12 people [on the jury].” The trial court’s determinations as affirmed by the California Supreme Court are presumed correct. Rice, 546 U.S. at 338-39, 126 S.Ct. 969 (“State-court factual findings, moreover, are presumed correct; the petitioner has the burden of rebutting the presumption by ‘clear and convincing evidence.’ § 2254(e)(1).”).
The majority’s expressed concerns about Olanders D.’s recusal are far from compelling. It is hardly surprising that a number of potential jurors expressed ambivalence about the death penalty. The fact that a prosecutor is more concerned with one potential juror’s ambivalence than another is not necessarily a sign of racial prejudice. Similarly, the fact that the majority in reviewing the voir dire transcripts thinks that a seated juror’s responses were no more responsive than Olanders D.’s is really of little moment. As noted, the trial judge—who heard Olanders D.’s voir dire—agreed with the prosecutor that he “demonstrated a lack of ability to express himself well.” The majority’s supposition that Olanders D.’s questionnaire responses *886may not have been “poor” is not clear or convincing evidence of anything. At most, the majority’s arguments and assumptions may suggest that the prosecutor’s evaluation of Olanders D. was not compelled, but none of them really question the sincerity of the prosecutor’s reasons or suggest a likelihood of some unstated improper motive. The majority fails to show that a fairminded jurist could not have agreed with the California Supreme Court.
The only indicia of possible racial bias was the fact that seven of the eighteen peremptory challenges exercised by the prosecutor excused African-American and Hispanic jurors. If this were enough to compel a finding of racial bias, there would be no reason for the second and third steps in the Batson standard or for deference to the trial court’s determinations. The lack of any compelling evidence of racial bias is clear when the record in this case is compared to the prosecutor’s statements in Kesser, 465 F.3d 351. There, in overcoming the deference due to the state court’s determinations, we commented: “The racial animus behind the prosecutor’s strike is clear. When he was asked to explain why he used a peremptory challenge to eliminate [a juror], he answered using blatant racial and cultural stereotypes.” Id. at 357. Here, in contrast, all the majority can do is suggest that other jurors, like Olanders D., were uncomfortable with the death penalty, failed to offer thoughtful answers, and did not communicate well. But even if the prosecutor’s perceptions about Olanders D. were incorrect or not unique, that fact would not be such compelling evidence of pretext as to justify a failure to defer to the California Supreme Court’s reasoned determination that the jurors were excused for proper, race-neutral reasons.
2. Gerardo 0.
Gerardo 0. was one of the recusals that Ayala challenged in his second objection. The prosecutor explained his challenge to Gerardo 0. as follows:
I made an observation of [Gerardo] when he first entered the courtroom on the first day that the jurors were called into the area.
At that time, he appeared to not fit in with anyone else. He was a standoffish type of individual. His dress and his mannerisms I felt were not in keeping with the other jurors.
He indicated to us at the beginning that he was illiterate. Actually, his words were that he was illiterate, and that he therefore had the questionnaire translated to him, so that he could make responses.
I observed him on subsequent occasions when he came to the court, and observed that he did not appear to be socializing or mixing with any of the other jurors, and I also take into account his responses on the questionnaire and in the Ho-vey questioning process, at which time he expressed that he had no feeling with regard to the death penalty in writing. When being questioned, he said that he was not sure if he could take someone’s life, or if he could take someone’s life into his hands.
He further responded in the Hovey process that there would be eleven other people, that he felt a little shaky as far as his responsibilities in this case.
For those reasons, I felt that he would be an inappropriate juror, and for that reason, I exercised the peremptory challenge.
The trial court accepted the prosecutor’s reasons. It noted that the record supported the prosecutor’s observations and commented that the recusal was based on Gerardo O.’s individual traits. The California Supreme Court in rejecting Ayala’s Wheeler/Batson claim noted that “Gerardo *887O. struggled with English and did not understand the proceedings.” Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 206.
The majority does not deny that Gerardo 0. stated that he was illiterate, or that he needed someone to fill out his questionnaire, or that he dressed differently, or that he did not mix with the other jurors. See Majority at pp. 854-57. Instead, the majority speculates that Ayala’s lawyer might have shown that despite his own comments, Gerardo 0. was not illiterate, and that Geraldo O.’s “dress and mannerisms were distinctly Hispanic.”17 Majority at p. 855. It further muses that the prosecutor’s comments concerning Gerardo O.’s manner and aloofness and his ambivalence toward the death penalty could have been pretexts for an underlying racial bias.18 Majority at pp. 856-57. Of course, it is impossible to negate such possibilities, but there is nothing but the majority’s imagination to fuel its assertions.19 Here, the trial judge agreed with the prosecutor’s observations of Gerardo 0. and the California Supreme Court affirmed. The majority has not presented the type of clear evidence that the United States Supreme Court has held is necessary to overcome our deference to state court findings. See Rice, 546 U.S. at 338-39, 126 S.Ct. 969.
3. Robert M.
Robert M. was one of the last persons whose recusal was challenged. The prosecutor explained his reasons as follows:
As far as [Robert MJ is concerned, Miss Michaels and I had discussions during the selection process here in court, even as late as immediately before the exercise of the last challenge.
The court would note that I had passed at one point, leaving [Robert M.] on. I have always felt some degree of reluctance with regard to [Robert M.], and my concern primarily is in the area of whether, after conviction, [Robert M.] would actually vote for the death penalty, and it was my view that taking all of his responses in Hovey into account, and *888the—some of his responses even as late as yesterday—for example, the following of the Sagon Penn case. It was Miss Michaels doing the questioning at that time, and I did not actually—it would have been possibly a disadvantage or a disservice to inquire further as to his impressions about the Sagon Penn case. I’m concerned about that case because the fact that Mr. Penn, in a very notorious trial here, was found not guilty in a second trial, and allegations of misconduct with regard to the District Attorney’s office and the police were certainly rampant in that case.
There’s really no way for me to inquire as to where [Robert M.] actually stood. As far as [Robert MJ is concerned, our scores, a combination of all the factors— Mr. Cameron graded [Robert M.] as a four, Miss Michaels had rated [Robert M.] as a five, and my score on him was four to a five, somewhere in that area. I had before doing any of the selection process, resolved that at the very best, we would not wish to have any jurors on this case whose combined score was five or less.
In spite of that, I passed once on him, but it is my view, basically, that because of his attitudes with regard to the death penalty, such as in his first response to whether he would always vote for—well, in the question number one about whether he would always vote for guilt, he indicated that it was a difficult question.
He said that he believed in the death penalty, but it was hard for him to be involved in the death penalty.
With regard to questions about whether he would vote for death, he said no, it would be hard to say, no, I don’t know what the evidence is, and Miss Michael’s reasons, which she expressed to me, and I have to agree with, is a great degree of concern about whether if we get to that point he could actually vote for death, and having that kind of a question in my mind as I’m trying this case would be distracting and worrisome to me during the process of the trial.
The trial judge accepted the prosecutor’s reasons, noting that although Robert M. “is certainly pro the death penalty,” his answers varied and “there may well be a legitimate concern as to whether or not he could impose it.” The court further noted that “an appropriate use of a peremptory would be for a person that any party feels either could not vote for death or could not vote for life.” In affirming Ayala’s conviction the California Supreme Court observed “that Robert M. was less than desirable from the prosecution’s point of view.” Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 206.
Again, the majority does not really question the prosecutor’s reasons, but speculates that had Ayala’s counsel been present he might have argued that Robert M.’s reluctance to impose the death penalty was not different from other jurors’ reluctance. Majority at pp. 857-58. In addition, the majority does not deny that Robert M. had stated that he had followed the Sagon Penn case, but argues that he only mentioned this briefly.20 Majority at p. 858. *889Nonetheless, Robert M.’s interest in a recent notorious criminal case that involved misconduct by the prosecutor and resulted in a not guilty verdict is a legitimate nondiscriminatory reason for recusal by the prosecutor.
4. The Other Jurors
The only other recused juror that the majority mentions is George S. See Majority at pp. 852-53 n. 16. The prosecutor explained that he had recused George S. because he (a) stated that he had sat on a prior jury and “was the one hold-out with regard to whatever issue was being presented at that time”; (b) was equivocal on the death penalty, (c) had been rejected as a police officer candidate, and (d) placed undue emphasis on the Bible. The trial judge commented that the prosecutor’s observations were accurate. The majority does not deny that the prosecution offered these individualized grounds for the recu-sal. Instead, the majority dismisses the fact that George S. had been a holdout juror with the comment that it was a civil action and speculates that George S.’s alleged emphasis on the Bible “cannot be evaluated at all” because of the loss of the questionnaires. See Majority at pp. 852-53 n. 16. Again, perhaps the majority’s assertions suggest that the prosecutor’s views were not compelled, but they do not really undermine their reasonableness or sincerity.
There were valid nondiscriminatory grounds for recusing the remaining three minority jurors. Galileo S. was recused because he (a) displayed a non-conformist attitude to the justice system, (b) had more run-ins with the law than he admitted, and (c) had an attitude that might create alienation and hostility on the part of other jurors. Luis M. was challenged because he (a) expressed ambivalence on the death penalty, (b) had investigated the case on his own, and (c) left the military with a low rank suggesting some sort of misconduct or inability to perform. Barbara S. was challenged because (a) her responses to oral questions were slow, (b) she had an empty look in her eyes and seemed out of tune with what was going on, and (c) her written and oral answers were incomplete and non-responsive.
As with the other recused jurors, the prosecution team offered individualized reasons for each of these recusals. There is no blatant racism, no reference to stereotypes (veiled or otherwise), and no dis-cernable pattern of discrimination in the reasons advanced by the prosecution. Nonetheless, these recusals are susceptible to the type of speculative challenges that the majority hurls at the recusals of Olan-ders D., Gerardo 0., and Robert M. In all likelihood, other jurors expressed ambivalence and equivalence about the death penalty, other jurors offered slow or incomplete responses, and other jurors probably had been denied employment or performed poorly in a job. These might be appropriate avenues to explore at the time that a recusal is made. But we are reviewing a 1989 state trial pursuant to AEDPA, and the Supreme Court in its recent opinions has reiterated that (a) Batson issues turn largely on evaluations of credibility, (b) the trial court’s determination is entitled to great deference, (c) the determination must be sustained unless it is clearly erroneous, and (d) AEDPA demands that state-court decisions be given the benefit of the doubt. See Felkner, 131 S.Ct. at 1307.
The California Supreme Court may not have been compelled to conclude that “the challenged jurors were excluded for prop*890er, race-neutral reasons.” Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 204. But its conclusion was objectively reasonable. That is, Ayala has not shown that the California Supreme Court’s ruling “was so lacking in justification that there was an error well understood and comprehended in existing law beyond any possibility for fairminded disagreement.” Richter, 131 S.Ct. at 786-87. As in Harrington, the majority’s opinion “illustrates a lack of deference to the state court’s determination and an improper intervention in state criminal processes, contrary to the purpose and mandate of AEDPA and to the now well-settled meaning and function of habeas corpus in the federal system.” Id. at 787.
IV
The Supreme Court decided Batson v. Kentucky in 1986, a year after Ayala killed three men and three years before his murder conviction. In Batson, the Supreme Court declined “to formulate particular procedures to be followed upon a defendant’s timely objection to a prosecutor’s challenges.” 476 U.S. at 99, 106 S.Ct. 1712. I would hold that Ayala’s claim that he had a constitutional right to have counsel present when the prosecutor offered its reasons for the challenged recusals was not dictated by precedent when Ayala’s conviction became final, and thus is Teag-ue-barred.
Ultimately, however, this case turns on the reasonableness of the California Supreme Court’s 2000 opinion that the absence of defense counsel and the loss of jury questionnaires were harmless error beyond a reasonable doubt as a matter of federal law. Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 204. Following the Supreme Court’s pointed guidance in Richter, 131 S.Ct. 770, and Williams, 133 S.Ct. 1088, we must conclude that the California Supreme Court adjudicated Ayala’s federal claims on their merits and thus apply AEDPA’s deferential standard of review.21
This standard of review mandates that we determine whether fairminded jurists could agree with the California Supreme Court. In other words, we can grant relief only if no fairminded jurist could find that the exclusion of defense counsel and the loss of questionnaires did not prevent Ayala from prevailing on his Batson claim. Here, the evidence of valid non-pretextual reasons for the prosecutor’s recusals renders the state court’s decision objectively reasonable.
Because the majority fails to appreciate that Ayala’s federal claim is Teague-barred, and applies a de novo standard of review, despite the Supreme Court’s contrary directions, I dissent.
. There does not appear to be any question that there was sufficient evidence to convict Ayala of murder. See Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 199-201.
. Because it is dear that Ayala's claims would be Teague-barred if reviewed de novo, the majority should have begun and ended with this Teague analysis, rather than reach the more difficult questions regarding the standard of review under 28 U.S.C. § 2254(d).
. In all quotations, the parallel citations have been omitted.
. The dissent argued that the majority’s choice of an adversarial proceeding over an in camera proceeding was contrary to the Supreme Court's decision in Batson not to formulate particular procedures. Id. at 1262 (Sneed, J., dissenting).
. The California Supreme Court carefully considered the prosecutor’s claim that his reasons for the recusals would disclose matters of strategy. It concluded that the prosecutor had “simply [given] the reasons for his challenges, reasons that defendant was entitled to hear and that disclosed no secrets of trial strategy.” Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 202-03. Accordingly, it concluded that "[i]t was unreasonable to exclude defendant from the hearings.” Id. 99 Cal.Rptr.2d 532, 6 P.3d at 203. The California Supreme Court was right. However, for purposes of the application of Teague, the point is that this conclusion was neither dictated nor compelled when Ayala’s conviction became final.
. As the majority notes, People v. Wheeler, 22 Cal.3d 258, 148 Cal.Rptr. 890, 583 P.2d 748 (1978), is the California analogue to Batson. See Majority at p. 836 n. 1. Accordingly, in cases arising out of California, claims that minority jurists were improperly excluded are sometimes referred to as Batson/Wheeler claims.
. This “alternate” approach, in addition to being incorrect, serves to obscure the majority's separation of the California Supreme Court’s determination that there was federal constitutional error from that court's determination that any error "was harmless beyond a reasonable doubt as a matter of federal law." Ayala, 99 Cal.Rptr.2d 532, 6 P.3d at 204 (internal citation omitted, emphasis added). Without its alternate approach, the majority would be hard pressed not to give the California Supreme Court’s determination of harmless error the deference it is entitled to under AEDPA. This, obviously, is our primary area of disagreement. I would, applying AEDPA, defer to the California Supreme Court’s opinion. The majority, instead, concocts an approach that circumvents AEDPA in order that it may review Ayala's 1989 conviction de novo. See Majority at p. 851 ("If we cannot say that the exclusion of defense counsel with or without the loss of the questionnaires likely did not prevent Ayala from prevailing on his Batson claim, then we must grant the writ.”).
. See Richter v. Hickman, 578 F.3d 944 (9th Cir.2009); Williams v. Cavazos, 646 F.3d 626 (9th Cir.2011).
. The panel had reasoned:
It is obvious, not "theoretical” or “speculative],” that Williams’s constitutional claim was not adjudicated at all, and so the Richter presumption is overcome. Id. at 785. Specifically, the portion of the court’s opinion concerning the discharge of Juror No. 6 reveals that the court upheld his dismissal on the sole basis that the trial court had not abused its discretion in applying section 1089. That the court engaged in an extended discussion of Williams’s stat*876utory claim, but made no mention whatsoever of her more fundamental constitutional claim, is a compelling "indication” that the court either overlooked or disregarded her Sixth Amendment claim entirely, rather than that it adjudicated the claim but offered no explanation at all for its decision.
Williams v. Cavazos, 646 F.3d 626, 639 (9th Cir.2011). The majority appears to follow a similar course in the case at bar.
. The Supreme Court specifically noted that "[i]n California, for example, the state constitutional right to be present at trial is generally coextensive with the protections of the Federal Constitution.” 133 S.Ct. at 1094-95 (internal quotation marks omitted).
. The majority states: "if we were compelled to determine whether the California Supreme Court adjudicated Ayala's federal claim on its merits in favor of the petitioner or the state, we would hold without the slightest hesitation that it found that error occurred under federal constitutional law.” Majority at p. 842.
. Section 2254(d) reads in relevant part:
(d) An application for a writ of habeas corpus on behalf of a person in custody pursuant to the judgment of a State court shall not be granted with respect to any claim that was adjudicated on the merits in State court proceedings unless the adjudication of the claim—
(1) resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States; or
(2) resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.
(e)(1) In a proceeding instituted by an application for a writ of habeas corpus by a person in custody pursuant to the judgment of a State court, a determination of a factual issue made by a State court shall be presumed to be correct. The applicant shall have the burden of rebutting the presumption of correctness by clear and convincing evidence.
. Although the majority here engages in considerably more analysis than we did in Felk-ner, the Supreme Court’s admonition remains instructive. The Court held:
The Batson issue before us turns largely on an "evaluation of credibility.” 476 U.S., at 98, n. 21, 106 S.Ct. 1712. The trial court’s determination is entitled to "great deference,” ibid.., and "must be sustained unless it is clearly erroneous,” Snyder v. Louisiana, 552 U.S. 472, 477, 128 S.Ct. 1203, 170 L.Ed.2d 175 (2008). That is the standard on direct review. On federal habeas review, AEDPA "imposes a highly deferential standard for evaluating state-court rulings" and "demands that state-court decisions be given the benefit of the doubt.” Renico v. Lett, 559 U.S. 766, 130 S.Ct. 1855, 1862, 176 L.Ed.2d 678, (2010) (internal quotation marks omitted). Here the trial court credited the prosecutor’s race-neutral explanations, and the California Court of Appeal carefully reviewed the record at some length in upholding the trial court’s findings. The state appellate court’s decision was plainly not unreasonable. There was simply no basis for the Ninth Circuit to reach the opposite conclusion, particularly in such a dismissive manner.
131 S.Ct. at 1307.
. The majority states:
In holding that Ayala has demonstrated his entitlement to relief under Brecht, we therefore also hold to be an unreasonable application of Chapman the California Supreme Court’s conclusion that Ayala was not prejudiced by the exclusion of the defense during Batson steps two and three or by the loss of the questionnaires.
Majority at p. 849 n. 12.
. This approach is also consistent with the Supreme Court's opinion in Cullen v. Pinhol-ster, - U.S. -, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011). There the Court held that "[i]f a claim has been adjudicated on the merits by a state court, a federal habeas petition must overcome the limitation of § 2254(d)(1) on the record that was before that state court.” Id. at 1400. The federal court must look at the record that was before the state court and determine whether fair-minded jurists could disagree with the state court's decision. Furthermore, the Court in reviewing the habeas petition in Pinholster applied a deferential standard of review. It held that "Pinholster has not shown that the California Supreme Court’s decision that he could not demonstrate deficient performance by his trial counsel necessarily involved an unreasonable application of federal law,” and that "Pinholster also has failed to show that the California Supreme Court must have unreasonably concluded that Pinholster was not prejudiced.” Id. at 1403-04, 1408.
. We further held that the right to a comparative juror analysis explicitly set forth in Miller-El was not Teague-barred as it "simply illustrates the means by which a petitioner can establish, and should be allowed to estab*882lish, a Batson error.” Boyd, 467 F.3d at 1146 (internal citation omitted).
.The majority’s cited quote from Hernandez v. New York, 500 U.S. 352, 111 S.Ct. 1859, 114 L.Ed.2d 395 (1991), demonstrates that Hernandez is not applicable to this case. The Supreme Court noted "the prosecutor’s frank admission that his ground for excusing these jurors related to their ability to speak and understand Spanish raised a plausible, though not a necessary, inference that language might be a pretext for what in fact were race-based peremptory challenges.” Id. at 363, 111 S.Ct. 1859. Here, the prosecutor did not mention any concern with Gerardo O.'s ability to speak Spanish and there does not appear to be any indication that any juror’s ability to speak Spanish was an issue. Instead, the majority, having poured over the record to determine that Gerardo O., despite his own admission of illiteracy, had "attended college in the United States,” opines that he "was perfectly capable of reading the summary of legal issues that was given to prospective jurors.” Majority at p. 855. It then leaps to the unsupported conclusion that the prosecutor’s purported reason for striking Gerardo O. was directly related to his status as someone who spoke Spanish as his first language. Majority at p. 855. The majority's speculation may not be illogical, but it is far from compelling.
. The majority also suggests that Gerardo O.’s ambivalence to the death penalty was no more pronounced than some seated white jurors. Majority at pp. 856-57. As previously noted, the potential jurors’ attitudes toward the death penalty was an important consideration for both the defense and the prosecution. The fact that the prosecutor distinguished between levels of ambivalence that the majority over twenty years later argues are indistinguishable is hardly a sign of pretext. Moreover, there is no doubt that Gerardo O.’s qualifications—professed illiteracy, distinctive dress and aloofness, and ambivalence to the death penalty—were unique.
. The majority goes so far as to fantasize that “perhaps” unbeknownst to the trial judge, Gerardo O. "had even organized a dinner for some of them at his favorite Mexican restaurant.” Majority at p. 856.
. The extent of the majority’s speculation is illustrated by its argument that because another juror who was seated mentioned that he was aware of the capital case People v. Harris, 28 Cal.3d 935, 171 Cal.Rptr. 679, 623 P.2d 240 (1981), the prosecutor's concern with Robert M.'s interest in the Sagon Penn case may have been pretextual. Majority at p. 858. This argument assumes that somehow the Harris case was similar to the Sagon Penn case. This seems unlikely as the crime in Harris took place in 1978, some eleven years before the jury selection process in this case. Moreover, unlike the alleged verdict in the *889Sagon Penn case, Harris was found guilty and the California Supreme Court’s opinion, which issued in 1981, did not find any serious misconduct by the district attorney.
. I do not agree with some of the majority's characterizations of my dissent. I have set forth my reasons in this dissent and trust the reader will be able to discern the respective merits of the majority and dissent without further assistance. To the extent the majority accuses me of relying heavily on recent Supreme Court opinions such as Richter, 131 S.Ct. 770, see Majority at pp. 850-51, n. 13, the accusation is accurate.