dissenting.
In defending a murder charge, it is a bad idea, I think, to leave the judge with a smoldering suspicion that your client had a role in killing two of the prosecution’s key witnesses before trial. A lawyer who minimizes that danger — by consenting to, rather than fighting, closure of the courtroom during the testimony of three surviving witnesses, out of a total of 18 testifying witnesses in the case — does not thereby render constitutionally ineffective assistance of counsel. That is all the more true, in my opinion, when there is not a shred of evidence that the closure had the slightest effect on the trial’s outcome.
The majority concludes otherwise. The majority suggests that the closure fight here was one worth having — indeed, that it was constitutionally mandated — its ethereal upside and concrete downside notwithstanding. The majority then sidesteps the absence of any actual prejudice resulting from Johnson’s consent to the closure— and with it, the plainly stated actual-prejudice requirement of Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) — by holding that Johnson need not show any prejudice at all in support of his ineffective-assistance-of-counsel claim. That holding is not supported by “clearly established Federal law, as determined by the Supreme Court,” 28 U.S.C. § 2254(d)(1), which means that we are without power to impose it on the Michigan state courts in this habeas case. I respectfully dissent.
* * *
The majority and I agree that Johnson’s habeas petition fails unless he can prove his ineffective-assistance-of-counsel claim. To prove that claim, Johnson must first show that his “counsel’s errors were so serious as to deprive [him] of a fair trial, a trial whose result is reliable.” Strickland, 466 U.S. at 687, 104 S.Ct. 2052. That showing requires, of course, proof that counsel committed errors in the first place. Second, Johnson “must show that there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Id. at 694, 104 S.Ct. 2052. We review for an abuse of discretion a district court’s decision whether to hold an evidentiary hearing on these issues. See Ivory v. Jackson, 509 F.3d 284, 297 (6th Cir.2007).
The district court did not abuse its discretion here, because Johnson’s claim clearly fails on both grounds. First, prior to trial, two witnesses identified Johnson as the shooter in a barrage that left Carlos Davis dead, Larry Lewis and James Mathis wounded, and Robert Richards untouched by the bullets that Johnson fired toward him. Richards was unavailable to testify at trial — because he had been shot to death, in bed, after his testimony at Johnson’s preliminary exam. Another prosecution witness, Elvin Robinson, was shot to death before he was able to testify at Johnson’s exam. At trial, the prosecution moved to close the courtroom during the testimony of three surviving witnesses: Damon Ramsuer, who was the sole surviving witness to have identified Johnson as the shooter; and Lewis and Mathis, two of the persons Johnson shot. In support of *449its motion, the prosecution explained that all three witnesses were “literally terrified” and were “cowering, refusing to come to court, even under threat of being arrested, because they [were] afraid for their lives.” Johnson’s counsel chose not to pick this particular fight — truly a strategic decision if there ever was one — and, after consulting with Johnson, affirmatively agreed to the three-witness closure.
The question before us is whether that decision was so far outside the bounds of competent representation as to amount to constitutionally ineffective assistance of counsel. I do not think the decision can possibly be seen that way. Having reviewed the trial transcript, it seems to me instead that the decision was correct. Johnson’s counsel essentially agreed to close the courtroom during the testimony of three witnesses, out of a total of 18 witnesses at trial. Strategically the net effect of that closure, as Johnson now describes it, was that several of his “female relatives” did not witness the testimony of those three witnesses. In return, Johnson’s counsel deflected the trial judge from a line of inquiry — as to why, exactly, these three witnesses were so terrified to testify against Johnson — that almost certainly would have reflected poorly on his client. That avoidance appears all the wiser given that it emerged at Johnson’s sentencing— and perhaps could have emerged sooner, had Johnson’s counsel fought the closure— that Richards was shot to death with ammunition from the same lot that Johnson used to kill Davis and wound Lewis and Mathis. The Constitution, suffice it to say, permitted this strategic choice.
Second, there is no evidence in the record — which includes the entire transcript of Johnson’s trial — that Johnson’s consent to closure as to three witnesses had any effect on the outcome of his murder trial. The majority does not assert the contrary. Instead, the majority holds, “[b]ecause the right to a public trial is a structural guarantee, if the closure were unjustified or broader than necessary, prejudice would be presumed.” Maj. Op. at 447. In so holding the majority drives right past the distinction between a Waller claim and a Strickland one. What the majority says is true enough for a Waller claim, but Johnson’s petition undisputedly turns on a Strickland one; and Strickland repeatedly and unequivocally says that actual prejudice is required. See, e.g., 466 U.S. at 694, 104 S.Ct. 2052 (“The defendant must show that there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different ”) (emphasis added); id. at 693, 104 S.Ct. 2052 (“It is not enough for the defendant to show that the errors had some conceivable effect on the outcome of the proceeding”).
Reasonable jurists can disagree as to whether, when a defendant asserts an ineffective-assistance claim based on an underlying violation of his right to a public trial, the Waller definition of prejudice should trump the Strickland one, or vice versa. In a lengthy analysis of this very issue, the Eleventh Circuit held that the Strickland actual-prejudice requirement applies. See Purvis v. Crosby, 451 F.3d 734, 743 (11th Cir.2006). (Our decision today directly conflicts with that decision.) The First Circuit has held that the Waller definition prevails, see Owens v. United States, 483 F.3d 48, 64-66 & n. 14 (1st Cir.2007)— although, critically, it did so in considering a challenge to an underlying federal conviction, see 28 U.S.C. § 2255, and thus was not limited, as we are here, to deciding whether “clearly established Federal law, as determined by the Supreme Court,” requires that result. Id. § 2254(d)(1) (emphasis added). But I simply do not see how, when Johnson presents a Strickland claim and Strickland by its terms imposes *450an actual-prejudice standard, we can hold that clearly established Supreme Court precedent required the Michigan state courts to apply a presumed-prejudice standard instead.
What has been clearly established for decades is that “a court must indulge a strong presumption that counsel’s conduct falls within the wide range of reasonable professional assistance[.]” Strickland, 466 U.S. at 689, 104 S.Ct. 2052. Johnson has not touched that presumption here. I respectfully dissent.