OPINION
ROGERS, Circuit Judge.
This case involves Antonio Garcia-Dorantes’ procedurally defaulted claim challenging the constitutionality of the jury selection computer program in Kent County, Michigan. In September 2001, a jury convicted Garcia-Dorantes of murder in the second degree and assault with intent to do great bodily harm less than the crime of murder. Before his trial started, Garcia-Dorantes failed to raise a Sixth Amendment challenge to the racial composition of the jury venire. In July 2002, the Grand Rapids Press published a story about a computer glitch in the Kent County software that had systematically excluded African-Americans from the jury pool from April 2001 through early 2002. As a result of the article, Garcia-Dorantes included a Sixth Amendment fair-cross-section claim in his direct appeal, which the state courts denied as procedurally defaulted due to his failure to object to the [588]*588composition of the jury venire at trial. Garcia-Dorantes then filed a 28 U.S.C. § 2254 habeas petition in federal court. The district court — attempting to apply the prejudice standard set forth in Ambrose v. Booker, 684 F.3d 638, 652 (6th Cir.2012), a companion case — found that Garcia-Dorantes had sufficiently demonstrated cause and actual prejudice to excuse his procedural default and had established a 'prima facie violation of his Sixth Amendment fair-cross-section right. The Respondent appeals. Because Garcia-Dorantes has shown cause and actual prejudice to excuse his procedural default and has established a prima facie violation of his Sixth Amendment fair-cross-section right, the district court properly granted habeas relief.
On September 10, 2001, a jury convicted Antonio Garcia-Dorantes of murder in the second degree and “assault with intent to do great bodily harm less than the crime of murder.” R. 29 (PgID 520). The district court described the offense and trial as follows:
[Garcia-Dorantes] was involved in a fight and stabbed two people, one fatally, in Grand Rapids, Michigan in the early morning hours of October 22, 2000. Jose Gomez, the homicide victim, died from a single stab wound to his upper chest, just beneath his collar bone. A 3-3/8 inch stab wound penetrated his left lung and punctured his pulmonary artery. The medical examiner determined that Gomez had been intoxicated at the time of the fight, with a blood alcohol level of .31 percent. Although the medical examiner testified that most persons would be comatose with a blood alcohol level above .30 percent, he could not discount the possibility that Gomez could have attacked someone in a fight situation if he was a habitual drinker.
Manuel Garcia was the assault victim. He worked for Gomez. The two men had been drinking that night until about 4:00 a.m. They arrived at Gomez’s house with a third companion, Gonzalo Ramirez-Toledo. When they drove up, [Garcia-Dorantes’] truck was parked across the street. [Garcia-Dorantes] testified that he thought the men were part of a gang and that they had tried to force him off the road during an earlier encounter that [day].1 [Garcia-Dorantes or one of the other men in his truck] was standing outside of his truck. [According to Garcia,] Gomez got out of Garcia’s truck and walked toward [Garcia-Dorantes], stating that he “did not want any problems”; he told the petitioner that he should leave or Gomez would call the police. Garcia and Ramirez-Toledo then got out of Garcia’s truck and approached [Garcia-Do-rantes], Two other persons had also exited [Garcia-Dorantes’] truck.' Ramirez-Toledo said [Garcia-Dorantes] had his hand behind his back when he first saw him standing beside his truck. Ramirez-Toledo said he did not see what occurred between [Garcia-Do-rantes] and Gomez, nor did he see a knife in [Garcia-Dorantes’] hand.
Garcia testified that when Gomez asked the men to leave, [Garcia-Dorantes] replied “And if I don’t want to?” and punched Gomez in the face. Gomez fell down, got up again and “threw himself’ at [Garcia-Dorantes]. Garcia acknowledged that he did not see which man [589]*589started the fight[, though he initially claimed that Garcia-Dorantes hit Gomez first]. As Garcia attempted to break the fight up, he felt a cramp in his leg and later learned that he had been stabbed in the buttock. Garcia was also stabbed in the back of his neck. Garcia testified that [Garcia-Dorantes] then threatened him with a bottle. Ramirez-Toledo testified that Gomez asked him to call the police. As Ramirez-Toledo did so, one of [Garcia-Dorantes’] friends hit him in the head with a bottle. [There is no record that Ramirez-Toledo ever connected with a 9-1-1 operator.] Ramirez-Toledo then heard Gomez say, “Let’s go.” Ramirez-Toledo observed [Garcia-Dorantes] get into his truck and leave the area, squealing his tires as he left. Gomez, Garcia, and Ramirez-Toledo [then] got into Garcia’s truck. Once inside, Garcia noticed that [Ramirez-Toledo] was very bloody. Garcia [left to drive] Gomez [and Ramirez-Toledo] to the hospital. [Garcia-Dorantes] and his companions had [already] left the area. While driving Gomez to the hospital, Garcia passed [Garcia-Dorantes’] truck when it stopped at a stop sign. When Garcia stopped at a red light, [Garcia-Dorantes] drove up and crashed into the rear of Garcia’s truck. Garcia continued driving but claimed that [Garcia-Do-rantes] crashed into his pickup truck two more times. Garcia drove to Gomez’s brother’s house, where they called the police and an ambulance. Police responded to a dispatch of a shooting to an address on Rose Street. Upon arrival, emergency personnel informed the police that it had actually been a stabbing and Gomez had died. Officers then received a dispatch for a “hit and run” and went to [Garcia-Dorantes’] house. The police were informed that [Garcia-Dorantes] had been involved in a hit and run accident. They noticed damage to the front of [Garcia-Dorantes’] vehicle. They arrested [Garcia-Dorantes] and his friend named Christian Diaz.
Police Officer John Riley testified that he interviewed [Garcia-Dorantes] in the early morning hours of October 22, 2000. Riley ascertained that [Garcia-Do-rantes] spoke very little English. Riley testified that he was “pretty fluent” in Spanish, and he read [Garcia-Dorantes] his Miranda rights in Spanish. Thereafter, [Garcia-Dorantes] made two verbal statements. In his first statement, he blamed his wife for the truck crash and resulting damage. Later that afternoon, he made a second statement in which he admitted to having been involved in a fight. [Garcia-Dorantes] told the police that he thought that Gomez and his friends were “gangbangers” who had threatened him earlier. Detective Gregory Griffin was present when both statements were made. He testified that although he found no evidence that any of the persons involved in this altercation were gang members, he could not rule out that Gomez was a gang member.
[Garcia-Dorantes] testified on his own behalf at trial, explaining that he was celebrating his daughter’s birthday on October 21, 2000. He and his friends later left the party to go to his girlfriend’s house, with whom he was having an extramarital affair. [Garcia-Do-rantes] parked in front of the house and went to the door. When there was no answer, he returned to his truck. It was then that he saw Garcia’s truck arrive. [Garcia Dorantes] claimed that Garcia had tried to ram him with his truck earlier that day and had tried to run [him] off the road. He thought Gomez, Garcia, and Toledo were gang members. According to [Garcia-Do-rantes], a person from the victim’s truck [590]*590provoked the fight. When others joined in, he became scared and pulled a knife, thrusting it once as a person lunged at him. It does not appear as though any of the other individuals had weapons. [Garcia-Dorantes] left and his friends followed. [Garcia-Dorantes] claimed that when he went to drive away;, the men pulled their truck in front of him and “locked” their brakes, causing [Gar-da-Dorantes’] vehicle to collide with their truck.
[Garcia-Dorantes] went home and told his wife about the accident. He said his wife volunteered to tell the police that she had been driving because [Garcia-Dorantes] was intoxicated at the time. [Garcia-Dorantes] denied intending to harm anyone.
[Garcia-Dorantes’] common-law wife, Anayeli Castellanos, testified that [Garcia-Dorantes] woke her in the early morning hours of October 22, 2000 and informed her that someone had crashed into his truck. [Garcia-Dorantes] demanded that she call the police to report the incident. The police arrived and arrested [Garcia-Dorantes]. Castella-nos admitted that she suggested that [Gracia-Dorantes] inform the police that she had been driving because [Garcia-Dorantes] appeared scared and had been drinking2.
[In closing, Garcia-Dorantes] argued that he acted in self-defense, and that the crime was no worse than manslaughter.3 [After the trial court instructed the jury on three separate, alternative charges related to Gomez’s death — murder in the first degree, murder in the second degree, and voluntary manslaughter — ][t]he jury found [Garcia-Do-rantes] guilty of the lesser offense of second-degree murder in Gomez’s death and assault with intent to do great bodily harm less than murder as to Garcia.
On September 30, 2002, following the conviction, the court sentenced Garcia-Do-rantes to fifteen to fifty years’ imprisonment on the second-degree murder charge, and five to ten years’ imprisonment on the assault charge, with the sentences to run concurrently.
On July 30, 2002, [after Garcia-Dorantes had been convicted,] the Grand Rapids Press reported that a computer glitch had [had] an impact on Kent County’s system for selecting jury venires. The glitch was introduced accidentally by the county when it assumed control of the jury selection computer program from a private vendor in April 2001. The problem came to light in 2002, when a local high school teacher, Wayne Bentley, completed a study of minority representation on Kent County juries. Bentley found that the underrepresentation of minorities was statistically significant, and shared his findings with county officials. The county subsequently conducted an internal study that revealed that “nearly 75 percent of the county’s 454,-000 eligible residents were excluded from potential jury pools since spring 2001” and that “[m]any blacks were excluded from ... jury pools due to a computer glitch that selected a majority of potential candidates from the suburbs.” The chief judge of the Kent County Circuit Court, George Buth, [591]*591stated, “There has been a mistake — a big mistake.”
Ambrose v. Booker, 684 F.3d 638, 640-41 (6th Cir.2012). In light of these revelations, Garcia-Dorantes — who had been convicted on September 10, 2001 — raised a Sixth Amendment fair-cross-section claim on direct appeal based on the computer glitch. People v. Garcia-Dorantes, No. 239306, 2003 WL 22416511, at *1-3 (Mich.Ct.App. Oct. 23, 2003). However, the Michigan Court of Appeals denied his claim, finding that Garcia-Dorantes had forfeited it when defense counsel failed to object to the jury array or panel. Id. at *2. The Michigan Supreme Court denied leave to appeal. People v. Garcia-Dorantes, 470 Mich. 865, 680 N.W.2d 897 (Mich.2004).
On July 12, 2005, Garcia-Dorantes filed a 28 U.S.C. § 2254 petition in district court, claiming, in part, that he had been denied his right to a fair trial because a computer glitch had resulted in the exclusion or underrepresentation of minorities in the jury venire. Garcia-Dorantes, No. 05-10172-BC, 2005 WL 2659056, at *1 (E.D.Mich. Oct. 18, 2005). The district court held the habeas proceeding in abeyance pending the exhaustion of state law remedies. Id. at *3. On February 15, 2007, the district court granted GarciaDorantes’ motion to reopen, Garcia-Dorantes v. Warren, No. 05-10172, 2007 WL 531094 (E.D.Mich. Feb. 15, 2007), and on March 8, 2011, denied all but Garcia-Dorantes’ Sixth Amendment claim. Garcia-Dorantes v. Warren, 769 F.Supp.2d 1092, 1112-13 (E.D.Mich.2011). The district court then ordered a magistrate judge to conduct an evidentiary hearing on the issue of whether Garcia-Dorantes “was deprived of his right under the Sixth Amendment to a trial by a jury chosen from a fair cross-section of the community.” Id. at 1112. The magistrate judge reviewed four items presented by stipulation: (1) the testimony of Wayne Bentley, a Grand Rapids school teacher and member of the Kent County jury commission. who uncovered the disparate representation; (2) the November 14, 2007 deposition of Terry Holtrop, the ease management manager for Kent County; (3) the report of Dr. Paul Stephenson; and (4) the report of Dr. Edward Rothman. This evidence is the same as that summarized in an earlier, companion case as follows:
First, the magistrate judge [reviewed] the testimony of Wayne Bentley, the teacher that uncovered the disparate representation. Second, the magistrate judge [considered the] testimony from Terry Holtrop, the case manager for the Kent County Circuit Court. Ambrose [v. Booker], 781 F.Supp.2d [532, 537-38 (E.D.Mich.2011) ]. Holtrop explained how Bentley’s evidence spurred an internal investigation by the County, culminating in the Kent County Jury Management System Report (hereafter “the Report”) dated August 1, 2002. The Report described how a transfer of control over a database — from a private vendor to the County in an effort to cut costs — caused the error:
[I]n the initial set-up of the Oracle database to accommodate the driver’s license and State ID data from the State file, an error was made in one parameter. Whether this was a programming error, the carry-over of a setting that existed within the Sybase database, misinterpreting instructions, or simply human error, that is now almost impossible to determine. The parameter that was entered within the database was 118,169. What should have been inserted within this setting was the total number of records in the State File, or 453,981 in 2001.
The net effect of this incorrect parameter is that the Jury Management [592]*592System performed a random selection against the first 118,169 jurors on the file. The percentage of jurors selected per Zip Code was proportional to the Zip Code composition of the first 118,169 records — but not Kent County as a whole. The total pool of prospective jurors from the State File is of course 3.8 times larger than the 118,-169 and hence the type of jury pool data as is evidenced in the various tables included in this report, for the second half of 2001 and the first half of 2002.
The next logical question being, why then did the jury pull from Zip Code 49341 jump so dramatically for 2001, from an average of 3.8% up to 10.24% ... and why did the jury pulls from Zip Code 49507 decline from an average of 8.56% to 2.13%.
The answer being that in 1998 (as was mentioned previously) the State File did not come in random order, but rather in Zip Code order ... lowest numbers to highest numbers. In subsequent years, new prospective jurors (either based on age or having moved to the County) were added to the end of the database. Existing prospective jurors (those that were on file the previous year) would simply have address information updated based on what the State provided. Their position in the dataset would not change. Therefore, the first 118,169 records of the dataset have a high percentage of lower numbered zip codes. As indicated on the map included in his packet, all the Zip Codes with the lower numbers are located outside of the Grand Rapids metro area.
Third, the magistrate judge considered statistical analysis submitted by Dr. Paul Stephenson, a statistician who used different methodologies to evaluate the impact of the glitch. Dr. Stephenson[, whose report focused primarily on January 2002,] first compared the percentage of eligible African-Americans in Kent County and the actual percentage in the venire. He found an absolute disparity of 6.03% and a comparative disparity of 73.1% fewer African-American members than would be expected. While Dr. Stephenson believed that these disparities were useful “descriptive statistics,” he did not believe they were “viable for inferential purposes.” Stephenson next used both a standard deviation and binomial test, and found that although the tests provided “insufficient evidence to demonstrate that the representation of Blacks or African-Americans in [a specific] venire is biased, this is in part due to the size of the venire.” Stephenson also noted [that] “while it would be less likely, one also could expect approximately 2% of all venires to contain no ... African-American members.” Stephenson then employed a “Chi-square Goodness-of-fit test.!’ Using this test, Dr. Stephenson found that “there is essentially no chance of acquiring the results we obtained if the selection process for potential jurors is unbiased. As a result, there is overwhelming evidence to conclude that the selection process for terms during the first months of 2002 was biased.” Dr. Stephenson concluded that “the analysis presented in this report demonstrates that a systematic bias did exist in the selection of individuals summoned for jury duty during the first three months of 2002. This bias would have inevitably led to under representation of ... African-Americans in the terms during this period of time.”
Fourth, the magistrate judge considered a report by Dr. Edward Rothman, who analyzed the composition of the Kent County Jury Pool between January 1998 and December 2002. Dr. Rothman used [593]*593the census figures from the 2000 census, at which time African-Americans comprised 8.24% of the population of Kent County[, and Hispanics comprised 5.98% of the population]. Dr. Rothman applied these figures to the period April 2001 to August 2002, and concluded that there was a 3.45% absolute disparity between jury-eligible African-Americans and those who appeared on jury veni-res[, and a 1.66% absolute disparity between jury-eligible Hispanics and those who appeared on jury venires]. Roth-man found a comparative disparity of 42% [for African-Americans and 27.64% for Hispanics].
Ambrose, 684 F.3d at 641-43. On January 5, 2012, the magistrate judge issued a recommendation that Garcia-Dorantes’ Sixth Amendment claim be allowed to proceed because his procedural default was excused4 and he had established a prima facie fair-cross-section violation.
Before the district court issued its opinion, we decided Ambrose v. Booker, 684 F.3d 638 (6th Cir.2012), a companion case arising from the same Kent County computer glitch, in which we held that a petitioner must show both cause and actual prejudice to excuse a procedural default, “even if the error structural.” Id. at 649. To guide district courts’ analyses of whether a petitioner has shown actual prejudice in a case alleging a violation of the Sixth Amendment fair-cross-section right, we explained:
We are then left with the question of the proper standard on remand. We are guided in part by the Eleventh Circuit’s analysis of a similar question in Hollis v. Davis, 941 F.2d 1471, 1480 (11th Cir.1991). In that case, the petitioner claimed that his counsel was ineffective for failing to object to Alabama’s systematic exclusion of African-American jurors from grand and petit juries. To excuse this default, the Holtis court required that petitioner show actual prejudice, which involved determining whether there was a reasonable probability that “a properly selected jury [would] have been less likely to convict.” Id. at 1482. The Eleventh Circuit’s analysis is persuasive. The most important aspect to the inquiry is the strength of the case against the defendant.5 As the Eleventh Circuit reasoned, “a transcript could [594]*594show a case against [petitioner] so strong, and defense so weak, that a court would consider it highly improbable that an unbiased jury could acquit.” Id. at 1483 (internal quotation marks omitted). In that circumstance, actual prejudice would not be shown.
Although the instant petitions do not involve a Strickland claim, this standard is appropriate because it balances the competing demands of constitutionally protected equal protection interests and comity toward the state courts. We recognize that the application of the actual prejudice standard in cases such as these presents a particularly challenging charge to the district courts below to answer the question, “what would have happened?” The law nonetheless requires that the question be answered— with a careful look at the transcripts involved, and with judgment that takes into account a fair balance of the competing interests of comity toward the final judgments of the state’s criminal processes and the protection of constitutional equal protection interests.
Ambrose, 684 F.3d at 652.
The district court, applying the actual prejudice standard set forth in Ambrose, conditionally granted Garcia-Dorantes’ petition for writ of habeas corpus. First, the district court found that Garcia-Dorantes had shown “cause and actual prejudice” to excuse his procedural default. In finding that Garcia-Dorantes had shown actual prejudice, the court began by rejecting Dr. Sommers’ expert testimony6 — testimony that posited that, in general, juries with more minority members are less likely to convict regardless of the crime — as irrelevant to answering the question of whether there is a reasonable probability that a properly selected jury would have been less likely to convict Garcia-Dorantes. The court explained:
[EJven if the Court credits the petitioner’s showing on this point as true [e.g., that juries with more minority members are less likely to convict], it is irrelevant to the question of actual prejudice. A properly selected jury could well have been all white, with no minority members at all. The petitioner is not entitled to a more lenient jury, or a jury panel with a particular racial balance— just one that has been selected through a constitutionally sound process, regardless of the race of the members. Taylor v. Louisiana, 419 U.S. 522, 538 [95 S.Ct. 692, 42 L.Ed.2d 690] (1975). The question is not whether the petitioner missed his chance to stand trial before a more merciful jury panel or a panel with a particular racial balance, but rather whether there is a reasonable probability that a different jury would have reached a different result.
The court then explained that even without Dr. Sommers’ expert testimony, a review [595]*595of Garcia-Dorantes’ trial record supported a finding that Garcia-Dorantes had shown actual prejudice:
Based on the record evidence, it is not reasonable to find that the “case against petitioner so strong, and defense so weak, that a court would consider it highly improbable that an unbiased jury could acquit.” Ambrose, 684 F.3d at 652.... There is no question that the trial record here discloses sufficient evidence on which a jury could have found that Garcia-Dorantes acted with malice and therefore reasonably could have found him guilty of second-degree murder. But “the question before [this Court] is not one of the sufficiency of the evidence.” Richey, 498 F.3d at 364. The evidence of the petitioner’s state of mind — his intent — was hardly overwhelming. [ ... ]
[T]he petitioner has raised a credible claim that on the facts of this case, “[a] mixed-race jury might clearly have a special perception.” Because African-American and Hispanic jurors from more urban areas would be more likely to have encountered gang violence than suburban white jurors, they may have better understood the situation that the petitioner faced in a 4:00 a.m. confrontation with a person who he thought was a gang member, and a jury selected from a fair cross-section of the Kent County community therefore might have a different perception of and reached different conclusions about the state of mind that the petitioner had during the resulting street fight. As the petitioner points out, none of the witnesses — including petitioner — testified to having a precise memory of what happened at the moment Jose Gomez was stabbed. Because the evidence on the petitioner’s state of mind reasonably allowed for competing inferences, the subjective perceptions, life experience, and common sense of jurors, as shaped by their individual racial and cultural backgrounds, would carry considerable weight in deciding what precise intent the defendant had at the crucial moment. That observation applies with equal force to the question whether the petitioner intended to inflict great bodily harm when he stabbed Manuel Garcia. Based on the facts and circumstances of the case, when comparing the result reached by a jury “selected by systematic exclusion of blacks, with the result which would have been reached by a racially mixed jury, [the Court] would have greater confidence in the latter outcome, finding much less probability that racial bias had affected it.” Ambrose, 684 F.3d at 652 n. 4.
Second, the district court found that Garcia-Dorantes had “established a prima facie violation of the fair-cross-section requirement under the Sixth Amendment,” showing both a systematic and substantial underrepresentation of minorities in jury venires. In evaluating the “fair and reasonable” representation prong from Du-ren, the court relied heavily on the magistrate judge’s recommendation:
Considering the absolute disparity figures cited by Dr. Rothman (3.45%) and Dr. Stephenson (6.03%), the magistrate judge noted that three recent cases have found an absolute disparity of 3% or less to be insufficient under Duren. United States v. Mujahid, 433 Fed.Appx. 559 (2011) (absolute disparity of 2.87% insufficient); United States v. Rodriguez-Lara, 421 F.3d 932 (9th Cir.2005) (3%); United States v. Royal, 174 F.3d 1 (1st Cir.1999) (2.9%). The magistrate judge also noted that neither absolute disparity figure would suffice under the standard requiring a minimum of 10% absolute disparity which has been adopted in other circuits, but he found that the 10% [596]*596rule was not dispositive because the Sixth Circuit has not adopted it.... That is ... reasonable ..., especially in light of the Supreme Court’s declination of the respondent’s invitation to adopt such an inflexible rule, finding no need in, such cases “to take sides today on the method or methods by which underrepresentation is appropriately measured.” Smith, 559 U.S. at 329-30 & n. 4, 130 S.Ct. 1382.
Considering the comparative disparity figures cited by Rothman (42%) and Stephenson (73.1%), the magistrate judge found that the latter figure would suffice under some recent holdings, but the former would not, citing United States v. Weaver, 267 F.3d 231 (3d Cir.2001) (comparative disparities of 40.01% and 72.98% insufficient, where absolute disparities were 1.23% and 0.71% respectively); United States v. Chanthadara, 230 F.3d 1237 (10th Cir.2000) (finding that comparative disparities of 40.89% and 58.39% were insufficient, but noting another holding that a 68.22% comparative disparity was sufficient); United States v. Clifford 640 F.2d 150 (8th Cir.1981) (46% comparative disparity insufficient).
After reviewing those authorities, the magistrate judge concluded that the evidence was sufficient to support a finding of underrepresentation under Duren, because it “presents a more obvious example of exclusion ‘inherent in the particular jury-selection process utilized’ than that in Smith.” The magistrate judge concluded that because the computer error “understated the county population by 3.8 times,” and because of the unique circumstance that the input file listed names in zip code order, where the lower zip codes represented suburban areas with much smaller proportions of African-American residents, it was apparent that the computer selection routine inherently and systematieally excluded African-Americans from the panels that were selected.
Respondent now appeals the district court’s grant of conditional habeas relief, claiming that (1) Garcia-Dorantes failed to show actual prejudice to excuse his procedural default, and (2) the district court erred in finding that the representation of African-Americans and Hispanics in the venire from which his jury was selected was not fair and reasonable.
As an initial matter, in determining whether to excuse Garcia-Dorantes’ procedural default, the district court applied the correct “actual prejudice” standard set forth in Ambrose v. Booker, 684 F.3d 638, 652 (2012), a standard intended to mirror the inquiry required by Strickland v. Washington, 466 U.S. 668, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Courts must consider whether, in light of the underrepresentation of African Americans in the jury venire, “there is a reasonable probability that ... the result of the proceeding would have been different.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. Stated another way, courts must ask, is there a reasonable probability that a different (i.e. properly selected) jury would have reached a different result, “a probability sufficient to undermine confidence in the outcome of the trial.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. Though Garcia-Dorantes argues that a less demanding “Hollis ” standard—which he contends asks only whether there is a “reasonable probability that a proper jury would have been less likely to convict,” Hollis v. Davis, 941 F.2d 1471, 1482 (11th Cir.1991)—applies, the Ambrose opinion stated clearly: “Although the instant petitions do not involve a Strickland claim, this standard is appropriate because it balances the competing demands of constitutionally protected equal protection in[597]*597terests and comity toward the state courts.” Ambrose, 684 F.3d at 652 (emphasis added). “This standard” refers to the Strickland standard — and not a different Hollis standard7 — a reading that wholly aligns with our previous statement that “federal courts should not reverse state court decisions unless a petitioner can show that the outcome would have been different.” Id. at 651 (emphasis added).
The district court also correctly found that Dr. Sommers’ expert testimony, in which Dr. Sommers stated that racially diverse juries are less likely to convict than all-white juries, was not relevant to the “actual prejudice” determination because his testimony: (1) does not support a finding that a different jury would have reached a different result; (2) lacks any individualized assessment of the case against Garcia-Dorantes; and (3) relies on impermissible racial stereotypes. First, Dr. Sommers’ testimony at best only supports a finding that a properly selected jury would have been less likely to convict — a standard that requires little more than a showing of a mere possibility of prejudice. This finding does not meet the more exacting Strickland standard, namely that Garcia-Dorantes had to show that, with a properly selected jury, there was a reasonable probability that the outcome of his trial “would have been different.” Ambrose, 684 F.3d at 651-52. To find that [598]*598there is a reasonable probability that with a properly selected jury a petitioner’s trial outcome would have been different, a court must do more than simply find that, in general, a more racially diverse jury is less likely to convict; such a finding says nothing about the case at hand. “To establish actual prejudice, a petitioner must show ‘not merely that the errors at his trial created a possibility of prejudice, but that they worked to his actual and substantial disadvantage, infecting his entire trial with error of constitutional dimensions.’ ” Hollis, 941 F.2d at 1480 (quotations and citation omitted).
Second, Dr. Sommers’ testimony lacked any individualized assessment of the evidence against Garcia-Dorantes, the factor we repeatedly identified in Ambrose as the “most important aspect to the inquiry.” Ambrose, 684 F.3d at 652. To permit a finding of actual prejudice based solely on such expert testimony would essentially eliminate the actual prejudice requirement in all but the most extreme cases (e.g., where the evidence against the defendant is so airtight that no reasonable jury could vote to acquit).
Third, reliance on the type of evidence provided by Dr. Sommers rests solely on general racial characteristics, considerations flatly inconsistent with the underlying theory of Batson v. Kentucky, 476 U.S. 79, 106 S.Ct. 1712, 90 L.Ed.2d 69, (1986) and its progeny. As Justice Marshall explained, in his Batson concurrence:
Exclusion of blacks from a jury, solely because of race, can no more be justified by a belief that blacks are less likely than whites to consider fairly or sympathetically the State’s case against a black defendant than it can be justified by the notion that blacks lack the “intelligence, experience, or moral integrity,” Neal v. Delaware, 103 U.S. 370, 394[, 13 Otto 370, 26 L.Ed. 567] (1881), to be entrusted with that role.
Batson, 476 U.S. at 104-05, 106 S.Ct. 1712 (Marshall, J., concurring).
Ultimately, the district court, applying the proper Strickland standard,8 correctly found that Garcia-Dorantes had shown actual prejudice to excuse his procedural default.9 A careful review of Garcia-Dorantes’ trial record indicates that there is a reasonable probability that a different jury would have reached a different result. Because Garcia-Dorantes conceded that he had; in fact, stabbed the victim, the trial focused on whether he had acted with the requisite intent. The jury thus was instructed on, and had a choice between, first-degree murder, second-degree murder, manslaughter, and self-defense and, as the district court explained, “[t]he evidence of [Garcia-Dorantes’] state of mind — his intent — was hardly overwhelming.” All of the witnesses, including Garcia-Dorantes, testified that they were (likely severely) intoxicated on the night of the fight, and none testified to having a precise memory of what happened at the moment Gomez was stabbed. Garcia-Do-rantes claimed that he had not intended to harm anyone, and that he had pulled a knife only after he became scared during [599]*599an attack by “gang members.” Individuals on both sides of the fight agreed that they did not really know each other prior to the fight. Though Garcia and Ramirez-Toledo both claimed that Garcia-Dorantes had provoked the fight and that none of the other fight participants had weapons, in what came down to a he-said-he-said scenario, it is far from clear that Garcia-Dorantes had intended to kill the victim. The trial court judge, at sentencing, even described the incident as “completely senseless,” stating, “I have to believe that perhaps a[n] element of uncontrolled machismo coupled with too much alcohol may have contributed to this.” Because there is a reasonable probability that a different jury — faced with the same evidence presented at Garcia-Dorantes’ trial' — would have returned a verdict on one of the alternative charges or acquitted Garcia-Dorantes based on self-defense, the district court correctly found that Garcia-Dorantes had shown actual prejudice.
We do not rely, however, on the district court’s reasoning that Garcia-Do-rantes “ha[d] raised a credible claim that on the facts of this case, ‘[a] mixed-race jury might clearly have a special perception.’ ” Though the court reasoned that, “[bjecause African-American and Hispanic jurors from more urban areas would be more likely to have encountered gang violence than suburban white jurors, they may have better understood the situation that the petitioner faced in a 4:00 a.m. confrontation with a person who he thought was a gang member ... [and thus have] reached different conclusions about [Garcia-Dorantes’] state of mind,” this argument fails for two reasons. First, as the Respondent points out, there is nothing inherent in race or ethnicity that would give a juror special insight into gang violence; rather, such an analysis appears more predicated on the experiences of jurors from urban and suburban settings. Since “[defendants are not entitled to a jury of any particular composition,” Taylor v. Louisiana, 419 U.S. 522, 538, 95 S.Ct. 692, 42 L.Ed.2d 690 (1975), it follows that they are not entitled to a jury predisposed to better “understand” their defense strategy.10 There is “no requirement that petit juries actually chosen must mirror the community and reflect the various distinctive groups in the population.” Id. at 538, 95 S.Ct. 692. As the district court explained, “[t]he question is not whether the petitioner missed his chance to stand trial before a more merciful jury panel or a panel with a particular racial balance, but rather whether there is a reasonable probability that a different jury would have reached a different result.” Second, this argument sounds much like the stereotyping arguments courts have sought to avoid, see Batson v. Kentucky, 476 U.S. 79, 104-05, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986) (Marshall, J., concurring), namely that because an individual is black or Hispanic, he is predisposed to better understand, or be sympathetic to, a defendant’s case. Ultimately, saying that a mixed-race jury might have a special perception (one born of differing experiences and circumstances) is not the same as saying that systematic underrepresentation in the jury venire led to an outcome much more likely to have been affected by racial bias — the harm the Sixth Amendment right guards against.
Finally, with respect to prejudice, the Respondent’s argument that Gar[600]*600cia-Dorantes cannot show prejudice because. he “cannot even show that the petit jury would more likely than not have been different,” incorrectly conflates the reasonable probability question with a preponderance of the evidence standard. Assuming arguendo that the Respondent’s statistical calculation is correct,11 a 48.51% likelihood that one of the two omitted African Americans would have been seated for Garcia-Dorantes’ petit jury likely satisfies the reasonable probability standard. Though a statistical showing that there is a tiny likelihood that the petit jury would have been different — absent the systematic underrepresentation — may support a finding of no actual prejudice, reasonable probability is not, either in law or in common parlance, limited to circumstances that are numerically greater than 50%.
In addition to showing cause and prejudice to excuse his procedural default, Garcia-Dorantes has established a prima facie violation of his Sixth Amendment fair-cross-section right because: (1) African-Americans and Hispanics are “distinctive” groups; (2) Respondent has forfeited any claim that the underrepresentation of minorities in the jury venire was not due to the systematic exclusion of the groups in the jury-selection process; and (3) the representation of the underrepresented groups in the jury venire was not “fair and reasonable.” Under Duren v. Missouri, 439 U.S. 357, 99 S.Ct. 664, 58 L.Ed.2d 579 (1979), to establish a prima facie case of a Sixth Amendment fair-
cross-section violation, a plaintiff must prove:
(1) that the group alleged to be excluded is a “distinctive” group in the community; (2) that the representation of this group in venires from which juries are selected is not fair and reasonable in relation to the number of such persons in the community; and (3) that this un-derrepresentation is due to systematic exclusion of the group in the jury-selection process.
Id. at 364, 99 S.Ct. 664. Because it is undisputed that Garcia-Dorantes has satisfied the first and third Duren prongs, we address only whether Garcia-Dorantes has shown that the representation of African Americans and Hispanics in the jury veni-re from which his jury was drawn was “not fair and reasonable in relation to the number of such persons in the community.”
Here, the absolute disparity for African-Americans of. 3.45% and corresponding 42% comparative disparity are sufficient to satisfy the Duren second prong. To determine whether the representation of a distinctive group is not “fair and reasonable;” courts typically use either — or both — the “absolute disparity” or the “comparative disparity” test to measure the distinctive group’s underrepresen-tation.
Absolute disparity measures the difference between the percentage of a group in the general population and its per[601]*601centage in the qualified wheel. For instance, if Asians constitute 10% of the general population and 5% of the qualified wheel, the absolute disparity is 5%. Comparative disparity measures the decreased likelihood that members of an underrepresented group will be called for jury service, in contrast to what their presence in the community suggests it should be. This figure is determined by dividing the absolute disparity of the group by that group’s percentage in the general population. In the example above, the comparative disparity is 50%: Asians are half as likely to be on venires as they would be if represented in proportion to their numbers in the community.
United States v. Shinault, 147 F.3d 1266, 1272 (10th Cir.1998).
Respondent argues that the disparities at issue here — 3.45% and 1.66% absolute disparities, and 42% and 27.64% comparative disparities for African Americans and Hispanics, respectively — fail to show that the representation of African Americans and Hispanics in the jury venires were not “fair and reasonable.” However, although Respondent points to numerous cases from other circuits in which courts have held that such numbers do not satisfy the Du-ren second prong, neither Supreme Court precedent, nor our prior decisions compel such a conclusion. First, the Supreme Court has not mandated the proper statistical measure to determine whether a minority is underrepresented for the purposes of Duren’s second prong. Berghuis v. Smith, 559 U.S. 314, 329, 130 S.Ct. 1382, 176 L.Ed.2d 249 (2010). In fact, in Smith, the Court explicitly refrained from “tak[ing] sides ... on the method or methods by which underrepresentation is appropriately measured.” Id. at 329-30, 130 S.Ct. 1382.
Second, we have previously found a 1.28% absolute disparity and 34% comparative disparity — disparity figures lower than those presented in this case — sufficient to satisfy the Duren second prong, particularly where the underrepresented population is a small percentage of the community. In doing so, we have emphasized that, “[wjhere the distinctive group alleged to have been underrepresented is small, ... the comparative disparity test is the more appropriate measure of under-representation.” Smith v. Berghuis, 543 F.3d 326, 338 (6th Cir.2008), rev’d on other grounds, 559 U.S. 314, 130 S.Ct. 1382, 176 L.Ed.2d 249 (2010). This is supported by the concern that absolute disparities “understate[] the systematic representative deficiencies in cases ... where ... the groups at issue comprise small percentages of the general population.” United States v. Weaver, 267 F.3d 231, 242 (3d Cir.2001) (quoting United States v. Shinault, 147 F.3d 1266, 1273 (10th Cir.1998)). For instance, in Smith, 7.28% of the Kent' County population was comprised of African-Americans eligible for jury service, yet only 6% of the venire panel members in Kent County were African-American. The Michigan Supreme Court determined that this disparity was constitutionally insignificant, but decided the case based on a failure to show the exclusion was “systematic.” Id.. at 339, 130 S.Ct. 1382. On appeal, we addressed the significance of the 1.28% absolute disparity and 34% comparative disparity.12 Id. at 336-39, 130 [602]*602S.Ct. 1382. Even though we recognized that a number of other courts had found absolute disparities in that range not to be constitutionally significant, id. at 337, 130 S.Ct. 1382 (collecting cases), we nevertheless found that it would be unwise to apply the absolute disparity test when the underrepresented group is so small; we concluded that in such situations, the comparative disparity test is more appropriate, reasoning:
“[A]n intractable use of the absolute measure may, in certain circumstances ... produce distorted results. For example, if a district with 10% non-white population has .5% non-whites in the wheel, the 9.5% disparity may not evoke disapproval under an absolute measure but may require it under a comparative measure.”
Id. at 337-38, 130 S.Ct. 1382 (quoting Foster v. Sparks, 506 F.2d 805, 835 (5th Cir.1975)). We then determined that, given the size of the minority population in Kent County, and the size of the comparative disparity (34%), the petitioner had satisfied the second prong of Duren.
The instant case is factually similar to Smith, a published precedent of our court. Both eases involve the jury pool in Kent County, a county in which African-Americans and Hispanics comprised a small percentage of the population, both in 1990 and 2000. Thus, as we found in Smith, the absolute disparity test — on its own — is of little use. Further, focusing on comparative disparity — as suggested by the Smith court — leads to the conclusion that Garcia-Dorantes has satisfied Duren’s second prong: the 42% comparative disparity here exceeds the 34% comparative disparity found sufficient in Smith. This finding comports with the finding of an Eighth Circuit court in United States v. Rogers, 73 F.3d 774 (8th Cir.1996). In Rogers, an Eighth Circuit panel indicated that a comparative disparity of 30% would be sufficient to meet the underrepresentation prong of the Duren test where African-Americans comprised only 1.87% of the jury-eligible population. Id. at 776-77. Though the Rogers court noted that it was bound by precedent that had previously upheld the challenged Iowa jury-selection plan, the majority urged the court to revisit the precedent given the 30% comparative disparity figure. The court reasoned that, because the minority population comprised only 1.87% of the total population, comparative disparity was the most accurate measure of the underrepresentation.13 Id.
Finally, the Ambrose district court, in its initial conditional grant of habeas, correctly considered and distinguished many of the cases from other circuits on which Respondent relies:
[T]he First Circuit and Second Circuit cases cited by Respondent for “similar” [603]*603statistics did not analyze a comparative disparity, see [United States v. Royal, 174 F.3d 1, 10 (1st Cir.1999) ], and [United States v. Rioux, 97 F.3d 648, 657-58 (2d Cir.1996) ], and the Third Circuit case rejecting comparative disparities of 40.01% and 72.98% was also faced with absolute disparities of only 1.23% and 0.71%, respectively, see [United States v. Weaver, 267 F.3d 231, 241-243 (3rd Cir.2001) ]. The absolute disparity in this case was much higher, at 3.45%. While the Seventh Circuit rejected an absolute disparity of 3%, the court noted at the same time that there was nothing to suggest that “this discrepancy amounts to anything more than a statistical coincidence.” [United States v. Ashley, 54 F.3d 311, 313-314 (7th Cir.1995).] Here, the discrepancy was indisputably caused by the error in compiling the jury pool. Indeed, Dr. Stephenson’s rejection of both the absolute and comparative disparity tests as statistically inadequate and his adoption of the Chi-square Goodness-of-fit test result as demonstrating the underrepresentation of African Americans in the jury pool is unrebutted.
Of the cases cited by Respondent in its objection, that leaves only the Tenth Circuit’s decision in [United States v. Orange], wherein the court rejected an absolute disparity of 3.57% and comparative disparity of 51.22%. [447 F.3d 792, 798-99 (10th Cir.2006) ]. The court relied on its earlier cases rejecting similar, and larger disparities. Id. However, a review of the Tenth Circuit cases does not reveal a persuasive rationale. Despite its reversal on other grounds, it is prudent to follow the Sixth Circuit’s decision in Smith.
Ambrose, 781 F.Supp.2d at 545. Given that African-Americans and Hispanics constitute a small percentage of the Kent County population, the absolute and comparative disparity figures here — taken together — are sufficient to satisfy Duren’s second prong.
Contrary to Respondent’s contention, in so finding, we have not “lowered” the standard for determining whether there is a “fair and reasonable” representation of a distinct group. See Appellant’s Br. at 51. Our circuit has never established a bright-line threshold for either absolute or comparative disparity figures, and in Smith, we clearly held that disparities lower than those present here satisfied the second prong of Duren. Nor do we agree with Respondent that finding that these disparity levels satisfy the Duren second prong requires an additional finding that there were “non-benign” factors at play. It is true that, in Smitk, we noted that “[ojther courts have ... found significant under-representation in the context of small minority populations that would otherwise not satisfy the requisite burden under Du-ren where ‘non-benign factors’ may be operating to produce such underrepresentation.” 543 F.3d at 339. However, we did not indicate that a showing of “non-benign” factors was always required for a petitioner to succeed on a claim where the relatively small size of the distinct group in the population makes the applicability of the absolute and comparative disparity tests problematic and, in any event, the 3.45% absolute disparity and 42% comparative disparity figures here are not so low as to require such an additional showing.
Because Garcia-Dorantes has satisfied the Duren second prong based on the combination of absolute and comparative disparity, we decline to apply in this case either the “disparity of risk” test14 or [604]*604“comparative disparity of risk”15 test discussed by Garcia-Dorantes. Neither Respondent nor Garcia-Dorantes argues for a wholesale adoption of either test, and both tests are relatively untested. We do not, however, reject the possibility that such tests may be useful in some future case.
Finally, because the glitch was inadvertent, no state interest was advanced by the computer error and subsequent underrepresentation of minorities in the jury venire. Once a petitioner establishes a prima facie violation of his Sixth Amendment fair-cross-section right — as Garcia-Dorantes has — the burden shifts to the government to show that “a significant state interest [is] manifestly and primarily advanced by those aspects of the jury-selection process ... that resulted] in the disproportionate exclusion of a distinctive group.” Duren, 439 U.S. at 367-68, 99 S.Ct. 664.
Accordingly, we affirm the judgment of the district court granting Gareia-Do-rantes’ petition for a writ of habeas corpus.