Garcia v. Quarterman

EDITH BROWN CLEMENT, Circuit Judge: In this habeas corpus action, the district court granted the petitioner relief. For the reasons that follow, we reverse. I. FACTS AND PROCEEDINGS Andrew Garcia (“Garcia”) and his brother, Alejandro Garcia (“Alejandro”), entered the car of a young woman while she was stopped at a traffic light in San Antonio, Texas. The Garcia brothers told the woman, at gun point, to drive to a secluded area and forced her to engage in various sex acts with both Alejandro and Garcia. In addition, the Garcia brothers used the victim’s debit card to withdraw funds from her account and forced her to purchase merchandise, which they kept. Eventually, the Garcia brothers released the victim. Garcia and Alejandro, each represented by separate counsel, were charged with aggravated sexual assault, aggravated robbery, and aggravated kidnapping. Both pleaded not guilty, and they were prosecuted in a single trial. The victim testified as the prosecution’s primary witness. Each brother also testified on his own behalf, with each implicating the other. 1 *443 Garcia essentially confirmed the victim’s version of events, except where her testimony implicated his willingness to commit the acts. Garcia’s theory of defense was that, while he committed the specific criminal acts alleged, he should not be convicted because he acted under duress. In support of this defense, Garcia testified that Alejandro threatened to kill Garcia and that “something would happen” to Garcia and his family if he did not cooperate with Alejandro. The jury was given two wholly-separate jury charges. Each charge was complete and self-contained on separate documents, and each charge was identified as applicable to each respective defendant. 2 Over the objection of the prosecution and Alejandro’s counsel, the trial judge included an accomplice-witness instruction in both Alejandro’s and Garcia’s jury charges. 3 The accomplice-witness instruction, given in Alejandro’s charge, described Garcia as an accomplice to Alejandro. And Garcia’s jury charge, in a similar instruction, described Alejandro as an accomplice to Garcia. Garcia’s jury charge also included instructions related to his duress defense. Garcia’s challenge is limited to the accomplice-witness instruction included in Alejandro’s jury charge. The jury convicted both Garcia and Alejandro on all counts. Garcia was sentenced to concurrent terms of sixty years for aggravated sexual assault, twenty years for aggravated robbery, and thirty-five years for aggravated kidnapping. Alejandro was sentenced to concurrent terms of seventy-five years for aggravated sexual assault, thirty-five years for aggravated robbery, and sixty years for aggravated kidnapping. Both Garcia and Alejandro appealed their convictions. 4 On direct appeal, Garcia raised only two issues, arguing that the trial court erred (1) by omitting Garcia’s requested jury instruction on the issue of the voluntariness of his conduct and (2) in restricting his cross-examination of the prosecution’s witnesses. The Texas court of appeals affirmed Garcia’s conviction. See Garcia v. State, No. 04-99-832-CR (Tex.App.—San Antonio Aug. 30, 2000). Garcia did not pursue any further direct review. Garcia filed a habeas petition in state court, alleging four grounds of error: (1) that his conviction was obtained pursuant to a constitutionally prohibited jury instruction; (2) that the jury instruction was harmful; (3) that he received ineffective assistance of counsel at trial; and (4) that he received ineffective assistance of counsel on appeal. The state trial court recommended granting habeas relief, but the Texas Court of Criminal Appeals disagreed and, without opinion, denied Garcia’s petition. Garcia next filed a habeas petition in federal court. Garcia’s petition focused on three issues: (1) that the inclusion of the accomplice-witness instructions in Alejandro’s jury charge violated the Fourteenth *444 Amendment; (2) that he received ineffective assistance of counsel at trial; and (3) that he received ineffective assistance of counsel on appeal. The district court initially dismissed Garcia’s petition as time-barred under 28 U.S.C. § 2244(d). On reconsideration, the district court equitably tolled the one-year statute of limitations and addressed the petition’s merits. 5 The district court concluded that Garcia’s conviction had been obtained pursuant to constitutionally prohibited jury instructions and that the state court’s denial of Garcia’s writ of habeas corpus was contrary to clearly established federal law as established by the Supreme Court. As a result, the district court granted relief by ordering Garcia’s convictions reversed and remanding to the state trial court for a new trial within 120 days of the district court’s order. The Director of the Texas Department of Criminal Justice (the “Director”), appeals. 6 II. STANDARD OF REVIEW “ ‘In a habeas corpus appeal, we review the district court’s findings of fact for clear error and review its conclusions of law de novo, applying the same standard of review to the state court’s decision as the district court.’ ” Martinez v. Johnson, 255 F.3d 229, 237 (5th Cir.2001) (quoting Thompson v. Cain, 161 F.3d 802, 805 (5th Cir.1998)). If the issue is a mixed question of law and fact, such as the assessment of harmless error, we review the district court’s determination de novo. Robertson v. Cain, 324 F.3d 297, 301 (5th Cir.2003) (citing Jones v. Cain, 227 F.3d 228, 230 (5th Cir.2000)). Garcia filed his habeas petition after the effective date of the Antiterrorism and Effective Death Penalty Act (“AEDPA”). As a result, the petition is subject to the procedures and standards imposed by AEDPA. See Lindh v. Murphy, 521 U.S. 320, 336, 117 S.Ct. 2059, 138 L.Ed.2d 481 (1997). Under AEDPA, “[a] federal court’s collateral review of a state-court decision must be consistent with the respect due state courts in our federal system.” Miller-El v. Cockrell, 537 U.S. 322, 340, 123 S.Ct. 1029, 154 L.Ed.2d 931 (2003). Moreover, our circuit precedent provides that “ ‘a federal habeas court is authorized by Section 2254(d) to review only a state court’s ‘decision,’ and not the written opinion explaining that decision.’ ” Pondexter v. Dretke, 346 F.3d 142, 148 (5th Cir.2003) (quoting Neal v. Puckett, 286 F.3d 230, 246 (5th Cir.2002) (en banc)). See also Santellan v. Cockrell, 271 F.3d 190, 193 (5th Cir.2001) (“The statute compels federal courts to review for reasonableness the state court’s ultimate decision, not every jot of its reasoning.”). A. Findings of Fact A state court’s factual findings are “presumed to be correct.” Hughes v. Dretke, 412 F.3d 582, 589 (citing 28 U.S.C. § 2254(e)(1)). Before a federal court, “a petitioner has the burden of rebutting this presumption with clear and convincing evidence.” Id. This deference extends not only to express findings of fact, but to the implicit findings of the state court. Sum *445 mers v. Dretke, 431 F.3d 861, 876 (5th Cir.2005); Young v. Dretke, 356 F.3d 616, 629 (5th Cir.2004). B. Conclusions of Law Our review of a state court’s conclusions of law is also deferential. The Supreme Court has determined that § 2254(d)(1) “affords a petitioner two avenues, ‘contrary to’ and ‘unreasonable application,’ ” by which to challenge a state court’s legal conclusions. Summers, 431 F.3d at 868-69 (discussing Williams v. Taylor, 529 U.S. 362, 405, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000)). A state court’s conclusion of law is “contrary to ... clearly established Federal law, as determined by the Supreme Court” under two conditions: (1) “the state court applies a rule that contradicts the governing law set forth in [the Supreme Court’s] cases,” or (2) “the state court confronts a set of facts that are materially indistinguishable from a decision of [the Supreme] Court and nevertheless arrives at a result different from [Supreme Court] precedent.” Foster v. Johnson, 293 F.3d 766, 776 (5th Cir.2002) (alteration in original) (quoting Williams, 529 U.S. at 405-06, 120 S.Ct. 1495). With respect to the second avenue, “a state court decision is an unreasonable application of clearly established Supreme Court precedent if the state court correctly identifies the governing legal rule but applies it unreasonably to the facts of a particular prisoner’s case.” Summers, 431 F.3d at 869 (internal quotations omitted). Federal habeas courts must heed the Supreme Court’s explanation that “ ‘unreasonable’ does not mean merely ‘incorrect;’ ” rather, “an application of clearly established Supreme Court precedent must be incorrect and unreasonable to warrant federal habeas relief.” Id. (quoting Foster, 293 F.3d at 776). “Only if a state court’s application of federal constitutional law fits within this paradigm may this court grant relief.” Id. III. DISCUSSION A. The accomplice-iuitness instruction Garcia contends that the trial court erroneously instructed the jury in violation of Sandstrom v. Montana, 442 U.S. 510, 99 S.Ct. 2450, 61 L.Ed.2d 39 (1979). 7 The accomplice-witness instruction in Alejandro’s jury charge, Garcia maintains, directed the jury to apply a mandatory conclusive presumption on an essential element of Garcia’s charged crime. 8 The instruction in Alejandro’s jury charge provided: You are instructed that an accomplice witness, as the term is hereinafter used, means any person connected with the crime charged, as a party thereto, and includes all persons who are connected with the crime, as such parties, by unlawful act or omission on their part transpiring either before or during the time of the commission of the offense and whether or not they were present and participated in the commission of the crime. The witness, Andrew Garcia, is an accomplice, if an offense was committed, and you cannot convict [Alejandro] upon *446 [Garcia’s] testimony unless you first believe that [Garcia’s] testimony is true and shows that [Alejandro] is guilty as charged, and then you cannot convict [Alejandro] upon said testimony unless you further believe that there is other testimony in the case, outside of the evidence of the said Andrew Garcia tending to connect [Alejandro] with the offense committed, if you find that an offense was committed, and the corroboration is not sufficient if it merely shows the commission of the offense, but it must tend to connect [Alejandro] with its commission, and then from all of the evidence you must believe beyond a reasonable doubt that [Alejandro] is guilty of the offense charged against him. Because the instruction defined Garcia as an accomplice, the argument continues, the trial judge negated the presumption of innocence and permitted the jury to convict Garcia without proof of guilt beyond a reasonable doubt as to every element of his charged crime. The district court agreed with Garcia and granted habeas relief. On appeal, the Director does not contest that the jury charge violated Sandstrom. Rather, the Director contends that a federal habeas court should review the trial court’s error for harmless error and that the district court erred in granting habeas relief without considering whether the Sandstrom error was harmless. We address each issue in turn. (1) Brecht harmless-error review is appropriate The Supreme Court has recognized two categories of constitutional violations, “trial error” and “structural defects.” See Arizona v. Fulminante, 499 U.S. 279, 307-08, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991). The former class is made up of those errors that “may be ‘quantitatively assessed in the context of other evidence presented.’ ” Myers v. Johnson, 76 F.3d 1330, 1337 (5th Cir.1996) (quoting Fulminante, 499 U.S. at 308, 111 S.Ct. 1246). Harmless error review is appropriate for the former class of errors but not the latter class. Id. Because the error alleged by Garcia falls into the former class, the error is subject to harmless error analysis. See Robertson, 324 F.3d at 304 n. 3 (“A Sandstrom-tjpe error has been held to be a ‘trial error’ to which the harmless error rule applies.”). Holding, as we do, that the complained-of error is subject to harmless error analysis does not complete our inquiry. In Brecht v. Abraliamson, 507 U.S. 619, 634-38, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993), the Supreme Court declined to apply the harmless-beyond-a-reasonable-doubt standard from Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), on habeas review of a state-court conviction. Rather, “[r]ecognizing the distinction between direct and collateral review,” the Court adopted the standard announced in Kotteakos v. United States, 328 U.S. 750, 776, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946). Brecht, 507 U.S. at 635, 638, 113 S.Ct. 1710. Under this Brecht harmless-error standard, the relevant question is whether the state court’s error “ ‘had substantial and injurious effect or influence in determining the jury’s verdict.’ ” Id. at 637, 113 S.Ct. 1710 (quoting Kotteakos, 328 U.S. at 776, 66 S.Ct. 1239). Following the enactment of AEDPA, some courts questioned whether federal habeas courts should continue to apply the Brecht harmless-error standard or whether the proper role was to review the state court’s application of the Chapman harm *447 less-error standard under AEDPA. 9 In this circuit — at least when the state court did not perform its own harmless-error review — we simply apply the Brecht harmless-error analysis. Robertson, 324 F.3d at 306 (“We hold that AEDPA’s restrictions on federal review of state habeas decisions do not alter Brecht’s mandate for harmless error analysis by federal courts when state courts have failed to address the question of harmless error.”). 10 Cf. Penry v. Johnson, 532 U.S. 782, 795-96, 121 S.Ct. 1910, 150 L.Ed.2d 9 (2001) (applying the Brecht harmless-error analysis in a post-AEDPA habeas proceeding where the state court found no error and, therefore, did not conduct a harmless-error inquiry). 11 And because here the state court did not perform a harmless-error analysis, it is enough for our purposes to follow Robertson. (2) The Sandstrom error was harmless The district court held that Garcia’s “Due Process rights were violated” because the “instructions relieved the State of its burden of proving by evidence Petitioner’s guilt in the alleged offense beyond a reasonable doubt.” In reaching this conclusion, the district court noted that the state trial habeas court recommended granting relief. In addition, after pointing out that Alejandro’s conviction had been overturned on direct review because of the erroneous inclusion of the accomplice-wit *448 ness instruction in his jury charge, the district court stated “that [Garcia] should be given the exact same treatment as his brother.” But the district court never assessed whether the Sandstrom error was harmless on the facts of Garcia’s conviction. The Supreme Court has had occasion to consider the application of harmless-error review to Sandstrom error. See Connecticut v. Johnson, 460 U.S. 73, 103 S.Ct. 969, 74 L.Ed.2d 823 (1983) (plurality). The Johnson plurality makes clear that, if a defendant concedes the issue on which the erroneous instruction wrongfully instructs the jury, Sandstrom error may be harmless. See Johnson, 460 U.S. at 87, 103 S.Ct. 969. Even the narrowest view, that of the Johnson plurality, described an example in which Sandstrom error would be harmless: “In presenting a defense such as alibi, insanity, or self-defense, a defendant may in some cases admit that the act alleged by the prosecution was intentional, thereby sufficiently reducing the likelihood that the jury applied the erroneous instruction as to permit the appellate court to consider the error harmless.” Id. In the Director’s view, Garcia’s case parallels Johnson’s example because Garcia’s entire defense was that, while he committed the specific acts alleged, he did so under duress. As such, the Director contends, the Sandstrom error was harmless. The Director also points out that the duress instructions, which were included in each count’s instructions for Garcia, ameliorate any persistent concern that the improper accomplice-witness instruction had a substantial and injurious effect on the jury’s verdict. We agree. At worst, the jury was instructed that Garcia was “connected with the crime charged, as a party thereto, ... by unlawful act or omission.” But Garcia admitted as much in establishing his own defense. By admitting his participation in the criminal acts, Garcia rendered the accomplice-witness instruction error harmless. Moreover, the accomplice-witness instruction did not affect Garcia’s duress defense adversely. The additional instruction given to the jury regarding Garcia’s duress defense eliminates any residual doubt as to the potential harm from the accomplice-witness instruction. The jury charge addressed the nature of Garcia’s defense: A defense set up by [Garcia] in this case is what is known as duress. It is an affirmative defense to prosecution that the actor engaged in the proscribed conduct because he was compelled to do so by threat of imminent death or serious bodily injury to himself or another. Compulsion within the meaning of this section exists only if the force or threat of force would render a person of reasonable firmness incapable of resisting the pressure .... Appearing in three separate sections of Garcia’s jury charge — one for each count against him — was an instruction on the application of Garcia’s duress defense: [I]f you find from the evidence ... that [Garcia] did commit the offense ... as alleged in the indictment and herein-before defined in this charge, but you further find ... that Alejandro Garcia did hold a firearm to the head of Andrew Garcia while threatening the said Andrew Garcia to commit the offense ... and that the threats of Alejandro Garcia were such threats of force as would render a person of reasonable firmness incapable of resisting the pressure, and that [Garcia] was in fear of imminent loss of life or serious bodily injury at the hands of Alejandro Garcia if he did not participate therein, then you will acquit [Garcia] and say by your verdict “not guilty.” *449 Viewing the jury instructions as a whole, as we must, see United States v. Young, 282 F.3d 349, 353 (5th Cir.2002), the resolution of the harmless error inquiry is clear. In error, the trial court instructed the jury that Garcia had performed the acts alleged. But the trial court correctly instructed the jury that it could not convict Garcia — even if he had engaged in the proscribed conduct — if the jury found he committed the acts under duress. The initial error was rendered harmless by the nature of Garcia’s duress defense. And the duress instructions properly stated Garcia’s defense. Garcia’s defense, which he reduced to the single question of duress, was rejected by the jury. Accordingly, we hold that the initial Sandstrom error did not have a “ ‘substantial and injurious effect or influence in determining the jury’s verdict.’ ” 12 Brecht, 507 U.S. at 637, 113 S.Ct. 1710 (quoting Kotteakos, 328 U.S. at 776, 66 S.Ct. 1239). See also Harris v. Warden, La. State Penitentiary, 152 F.3d 430, 439-40 (5th Cir.1998) (holding, in a pre-AEDPA case, erroneous jury instructions harmless under Brecht because the evidence and defense presented related to an issue unaffected by the error). B. Ineffective assistance of counsel In addition to the Sandstrom error claim, Garcia also asserted that he received ineffective assistance of counsel both at trial and on appeal under Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). 13 Garcia contends that counsel rendered constitutionally ineffective assistance (1) at trial by arguing in favor of the erroneous accomplice-witness instruction and (2) on appeal by failing to raise the Sandstrom error. These same Strickland claims were asserted in Garcia’s state habeas petition and summarily rejected by the state habeas court. And, though the state habeas court denied relief without comment, under 28 U.S.C. § 2254(d), we review only the state court’s ultimate decision. See Neal, 286 F.3d at 246. Our task then is to determine whether the state habeas court’s rejection of Garcia’s Strickland claim is contrary to or an unreasonable application of Supreme Court precedent. Before the state habeas court, Garcia was required to show (1) “that counsel’s performance was deficient” and (2) “that the deficient performance prejudiced the defense.” Strickland, 466 U.S. at 687, 104 S.Ct. 2052. To substantiate the deficient performance claim, Garcia was required to show that counsel’s actions “fell below an objective standard of reasonableness.” Id. at 688, 104 S.Ct. 2052. And to establish that counsel’s deficient performance prejudiced his defense, Garcia had to “show ‘that counsel’s errors were so serious as to deprive the defendant of a fair trial, a trial whose result is reliable.’ ” Lockhart v. Fretwell, 506 U.S. 364, 369, 113 S.Ct. 838, 122 L.Ed.2d 180 (1993) (quoting Strickland, 466 U.S. at 687, 104 S.Ct. 2052). *450 Our AEDPA review of the state habeas court’s rejection of these claims is simplified by our prior determination that the Sandstrom error was harmless with respect to Garcia. The state habeas court’s rejection of Garcia’s Strickland claim was not unreasonable under Strickland because the allegedly deficient performance of counsel at trial produced no harm to Garcia’s cause. See Gochicoa v. Johnson, 238 F.3d 278, 286 (5th Cir.2000) (finding no Strickland prejudice where counsel’s error was harmless); Mayabb v. Johnson, 168 F.3d 863, 869 (5th Cir.1999) (finding no Strickland prejudice where the argued-for “jury charge did not have a substantial and injurious effect or influence in determining the jury’s verdict”); Harris, 152 F.3d at 440 (finding no Strickland prejudice based on the court’s earlier conclusion that the erroneous jury instructions were harmless under Brecht). Nor is counsel’s failure to raise the Sandstrom error on appeal constitutionally deficient: “When we do not find prejudice from the trial error, by extension, we cannot find prejudice from an appellate error predicated on the same issue.” Mayabb, 168 F.3d at 869 (citing Ricalday v. Procunier, 736 F.2d 203, 208 (5th Cir.1984)). Moreover, contrasting the dissimilar results between Garcia and Alejandro does not establish prejudice. To the contrary, the disparity supports our analysis because Garcia, unlike Alejandro, conceded that he committed criminal acts. By contrast, Alejandro argued that the victim was a willing participant. It comes, then, as no surprise that the accomplice-witness instruction could be harmful — even reversible^ — error as to Alejandro but not as to Garcia. Accordingly, we hold that the state habeas court’s rejection of Garcia’s Strickland claim was neither contrary to nor an unreasonable application of Supreme Court precedent. IV. CONCLUSION We REVERSE the district court’s grant of habeas relief and RENDER judgment in favor of the Director. Footnotes [1] Testifying in narrative form, Alejandro contended that the victim had consented to engage in the sexual acts with him. He also stated that the victim had allowed him to *443 withdraw the money and had purchased the merchandise willingly in exchange for marijuana. Alejandro denied the presence of any firearm in the vehicle and claimed that Garcia forced the victim to perform similar acts on Garcia because she had performed them on Alejandro. [2] In addition, there were separate verdict forms (also on separate documents) as to each defendant. [3] Garcia’s attorney argued in favor of including the accomplice-witness instruction. [4] Alejandro's conviction was overturned on direct appeal because of the inclusion of the accomplice-witness instruction. See Garcia v. State, No. 04-99-897-CR (Tex.App.—San Antonio Oct. 18, 2000). [5] In addition to the appeal on the merits, the Director also appeals the district court’s decision to equitably toll the statute of limitations. Because we deny the habeas petition, we do not address the issue of tolling except to note that in this circuit an error of law, however reasonable it might be, " ‘is not an extraordinary circumstance such that equitable tolling ís justified.' " United States v. Riggs, 314 F.3d 796, 799 (5th Cir.2002) (quoting Cousin v. Lensing, 310 F.3d 843, 849 (5th Cir.2002)). [6] The district court granted the Director's motion to stay judgment pending the outcome of this appeal. [7] Garcia also cites Carella v. California, 491 U.S. 263, 109 S.Ct. 2419, 105 L.Ed.2d 218 (1989), Francis v. Franklin, 471 U.S. 307, 105 S.Ct. 1965, 85 L.Ed.2d 344 (1985), and Connecticut v. Johnson, 460 U.S. 73, 103 S.Ct. 969, 74 L.Ed.2d 823 (1983) (plurality), in support of his claim. [8] We assume without deciding that Alejandro’s jury charge can be prejudicial to Garcia. [9] Compare Gutierrez v. McGinnis, 389 F.3d 300, 306 (2d Cir.2004) (declining to apply Brecht "when a state court explicitly conducts harmless error review of a constitutional error [because] a habeas court must evaluate whether the state unreasonably applied Chapman”), with Webber v. Scott, 390 F.3d 1169, 1177 (10th Cir.2004) (applying Brecht when the state court did not apply the Chapman harmless-error standard), with Inthavong v. Lamarque, 420 F.3d 1055, 1059 (9th Cir.2005) (holding that a federal habeas court must perform both an AEDPA-based review of the state court's Chapman harmless-error analysis and a Brecht harmless-error review of the original error as a precondition to granting habeas relief). [10] In the time since Robertson was decided, the Supreme Court has not addressed the specific question of whether federal habeas courts are to apply Brecht after AEDPA's enactment. However, in Mitchell v. Esparza, the Supreme Court, on habeas review, considered the state court's harmless-error review under AEDPA’s standard of review: "We may not grant respondent's habeas petition, however, if the state court simply erred in concluding that the State’s errors were harmless; rather, habeas relief is appropriate only if the [state court] applied harmless-error review in an 'objectively unreasonable' manner.” 540 U.S. 12, 18, 124 S.Ct. 7, 157 L.Ed.2d 263 (2003) (per curiam) (citing Lockyer v. Andrade, 538 U.S. 63, 75-77, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003)). The Second Circuit has read Mitchell as implicitly rejecting Brecht, at least when the state court has made a harmless-error determination. Gutierrez v. McGinnis, 389 F.3d 300, 306 (2d Cir.2004). But even if we read Mitchell as the Second Circuit does, our precedent in Robertson remains unaffected because Robertson speaks directly to the situation where the state court did not perform a harmless-error review. Nor does our adherence to Robertson signal a rejection of the Ninth Circuit’s two-step harmless-error habeas review; under that analysis, a finding against the petitioner on either step precludes habeas relief. See Inthavong v. Lamarque, 420 F.3d 1055, 1059 (9th Cir.2005). [11] At oral argument, the Director suggested that state courts first should apply the Brecht harmless-error analysis, which federal habeas courts would then review under AEDPA's deferential standard. We disagree. Brecht’s less-stringent standard of harmless-error review is predicated on a federal habeas court's respect for comity, finality, and federalism. Congress addressed these same principles in enacting AEDPA. See Williams, 529 U.S. at 436, 120 S.Ct. 1479 ("AEDPA's purpose [is] to further the principles of comity, finality, and federalism.”). Brecht’s existence does not relieve a state court of its duty to apply Chapman's harmless-error analysis where it is required to do so under Supreme Court precedents. [12] In the alternative, we hold that the state habeas court’s rejection of Garcia's post-conviction challenge was not contrary to or an unreasonable application of Supreme Court precedent. Given the similarities between the record before us and the example in Johnson, it would not be unreasonable for the state habeas court to conclude that the Sandstrom error was harmless beyond a reasonable doubt. [13] The district court did not grant habeas relief based on the Strickland claim. Nevertheless, the district court determined that Garcia received ineffective assistance of counsel in relation to another issue below.