Joseph Howell v. Superintendent Rockview SCI

939 F.3d 260

• Court: Court of Appeals for the Third Circuit
• Decided: September 17, 2019
• Docket: 17-1758
• Status: Published

Opinion

PRECEDENTIAL

UNITED STATES COURT OF APPEALS FOR THE THIRD CIRCUIT ________________

No. 17-1758 ________________

JOSEPH HOWELL, Appellant

v.

SUPERINTENDENT ROCKVIEW SCI; ATTORNEY GENERAL PENNSYLVANIA; DISTRICT ATTORNEY ALLEGHENY COUNTY ________________

On Appeal from the United States District Court for the Western District of Pennsylvania (D.C. No. 2-12-cv-00884) District Judge: Honorable David S. Cercone ________________

Argued May 1, 2019 Before: RESTREPO, PORTER and FISHER, Circuit Judges. (Filed: September 17, 2019)

Leigh M. Skipper, Chief Federal Defender Helen Marino, First Assistant Federal Defender Arianna J. Freeman Loren D. Stewart [ARGUED] Federal Community Defender Eastern District of Pennsylvania Federal Community Defender Office for the Eastern District of Pennsylvania 601 Walnut Street The Curtis Center, Suite 540 West Philadelphia, PA 19106 Counsel for Appellant

Stephen A. Zappala, Jr., District Attorney Ronald M. Wabby, Jr., Deputy District Attorney Rusheen R. Pettit [ARGUED] Allegheny County Office of District Attorney 436 Grant Street Pittsburgh, PA 15219, Counsel for Appellees ________________

OPINION OF THE COURT ________________

FISHER, Circuit Judge. Criminal defendants are deprived of their Sixth

2 Amendment right to a jury selected from a broad representation of the community when distinctive groups are systematically excluded from the jury selection process. See Duren v. Missouri, 439 U.S. 357, 363-64 (1979). Because any under- representation in Joseph Howell’s jury pool was not caused by a systematically discriminatory process, the District Court properly denied his habeas petition alleging a Sixth Amendment violation. We will affirm. I. Jury selection in Howell’s 2004 prosecution consisted of two venire panels. The first included thirty-five individuals, two of whom were black but were both excused for hardship. The second panel included twenty-five potential jurors, all of whom were white. Ultimately, Howell, a black man, was convicted for the 2002 felony murder of a white man by an all- white jury. Prior to jury selection, Howell filed a Motion to Ensure Representative Venire, arguing that he was entitled to a jury pool that represented a fair cross section of the community— Allegheny County—particularly with respect to race. The trial court held a hearing on Howell’s allegations that black individuals were systemically under-represented in Allegheny County’s jury pools, during which it adopted the record from two other cases where defendants also raised a fair-cross- section challenge. The incorporated record included expert testimony from Dr. John F. Karns, a sociologist, regarding the racial statistics and demography of Allegheny County. Dr. Karns’ testimony expounded on demographic data gathered over a six-month period in 2001, over a ten-day period in 2002, and from the 2000 census. The 2001 study was based on data gathered by the firm Gentile Meinert & Associates and interpreted by Dr. Karns. Gentile Meinert &

3 Associates provided prospective jurors (individuals who appeared for jury selection pursuant to a summons) with a paper survey that asked questions about their race, age, and gender. From this study, which surveyed approximately 4500 potential jurors, Dr. Karns calculated that black individuals made up 4.87% of Allegheny County’s jury pool. He also found that black individuals made up 10.7% of the population of Allegheny County eligible for jury service. Based on these numbers, Dr. Karns concluded that “whites [were] overrepresented” in jury pools, resulting in systematic exclusion of “a significant number of people for a significant time.” App. at 112, 127. Despite this conclusion, the trial court denied Howell’s motion. An all-white jury was impaneled and found Howell guilty of felony murder. Howell moved for extraordinary relief, arguing that he should be retried by a representative jury, even if assembling the jury would require multiple venires. The trial court denied his motion; it then sentenced Howell to a mandatory sentence of life without parole. Howell timely appealed to the Pennsylvania Superior Court, which held that Howell had not been denied a trial by a fair cross-section of the community. The Superior Court noted Dr. Karns’ testimony,1 and identified the proper test for determining whether a fair-cross-section violation occurred. The court then concluded that Howell “fail[ed] to demonstrate ‘an actual discriminatory practice in the jury selection process,’” and, therefore, held that Howell did not demonstrate a constitutional violation. App. at 252-54 (quoting

1 The Superior Court observed Howell’s reliance on Dr. Karns’ testimony without stating whether it was reliable or making a finding of fact about its accuracy and declined to reach the statistical analysis.

4 Commonwealth v. Johnson, 838 A.2d 663, 682 (Pa. 2003)). The state court stated that, though the U.S. Supreme Court’s test does not require a showing of discriminatory intent, it was bound to follow Pennsylvania Supreme Court precedent, which does require such a showing. Howell filed a habeas petition based on six grounds, including his fair-cross-section claim. A magistrate judge issued a report and recommendation that assumed, without deciding, “that the Superior Court erred in requiring [Howell] to show discriminatory intent,” but concluded that, under de novo review, Howell failed to establish a Sixth Amendment violation. App. at 14-16. The magistrate judge compared the level of racial disparity in Howell’s case to those in other cases around the country. She concluded that, because other courts found no constitutional violation in cases with higher percentages of disparity than here, Howell could not establish his claim. The District Court adopted the magistrate judge’s report and recommendation and denied Howell’s petition. Howell now appeals. II. The District Court exercised subject matter jurisdiction pursuant to 28 U.S.C. §§ 2241 and 2254. We exercise appellate jurisdiction pursuant to 28 U.S.C. §§ 1291 and 2253. The District Court did not hold an evidentiary hearing but relied exclusively on the state court record; we therefore undertake a plenary review of the District Court’s order utilizing the same standard that the District Court applied. Branch v. Sweeney, 758 F.3d 226, 232 (3d Cir. 2014). The Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”) dictates the parameters of our review and

5 requires us to afford considerable deference to the state court’s legal and factual determinations. Lambert v. Blackwell, 387 F.3d 210, 234 (3d Cir. 2004). We may overturn a state-court holding only where it “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law,” or “was based on an unreasonable determination of the facts in light of the evidence presented.” Id. (quoting 28 U.S.C. § 2254(d)(1)-(2)). The state court’s factual conclusions “‘shall be presumed to be correct’ unless the petitioner rebuts ‘the presumption of correctness by clear and convincing evidence.’” Id. (quoting 28 U.S.C. § 2254(e)(1)). If the state court erred, habeas relief should be granted only if, upon de novo review, the prisoner has established that he “is in custody in violation of the Constitution or laws or treaties of the United States.” 28 U.S.C. § 2254(a); see also Saranchak v. Beard, 616 F.3d 292, 301 (3d Cir. 2010). III. The Sixth Amendment promises all criminal defendants a trial by a “jury drawn from a pool broadly representative of the community . . . as assurance of a diffused impartiality.” Taylor v. Louisiana, 419 U.S. 522, 530-31 (1975) (quoting Thiel v. S. Pac. Co., 328 U.S. 217, 227 (1946) (Frankfurter, J., dissenting)). A violation of this right occurs where “jury wheels, pools of names, panels, or venires from which juries are drawn . . . exclude distinctive groups in the community.” Duren, 439 U.S. at 363-64 (quoting Taylor, 419 U.S. at 538). Howell argues that his Sixth Amendment rights were violated by Allegheny County’s systematic exclusion of black jurors at the time of his trial. A. A state-court decision is “contrary to” or an

6 “unreasonable application of” federal law if it directly conflicts with Supreme Court precedent or reaches a different result than the Supreme Court when presented with materially indistinguishable facts. Williams v. Taylor, 529 U.S. 362, 405 (2000). In its analysis, the state court relied on its interpretation of Pennsylvania Supreme Court precedent to determine whether Howell established a prima facie violation of his right to a jury composed of a representative cross-section of his community. Quoting Commonwealth v. Estes, 851 A.2d 933 (Pa. Super. Ct. 2004) (citing Johnson, 838 A.2d 663), the court set forth the Duren standard for establishing such a violation— that (1) an allegedly excluded group is “distinctive” in the community; (2) the group’s representation in jury-selection panels is not fair and reasonable in relation to the community’s population; and (3) the group is under-represented due to its systematic exclusion from the jury-selection process—but then went on to state that “[p]roof is required of an actual discriminatory practice in the jury selection process, not merely underrepresentation of one particular group.” App. at 252-54. The state court acknowledged Howell’s argument that he was “not required to prove discriminatory intent . . . under Duren,” but the court concluded that “the Pennsylvania Supreme Court has held otherwise” and that it was “bound by [that] prior decision[].” App. at 253-54. Irrespective of how the Superior Court reached its conclusion, that conclusion must comport with “clearly established Federal law as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1); see also Williams, 529 U.S. at 412 (“As the statutory language makes clear . . . § 2254(d)(1) restricts the source of clearly established Federal law to [the Supreme] Court’s jurisprudence.”). Therefore, the question before us is whether the Superior

7 Court’s decision is consistent with Duren and its progeny. Duren established a three-factor test for determining when a fair-cross-section violation has occurred. Significantly, that test does not include a requirement for proof of discriminatory intent. To the contrary, the Court—in a footnote—distinguished the Sixth Amendment claim before it from cases brought under the Equal Protection Clause by noting that, in the latter, a showing of discriminatory purpose is essential, but that, in the former, “systematic disproportion itself demonstrates an infringement.” Duren, 439 U.S. at 368 n.26. The Commonwealth correctly notes that the Court’s statements in a footnote are not necessarily binding authority on habeas review because “‘clearly established Federal law’ . . . includes only the holdings, as opposed to the dicta, of [the] Court’s decisions.” Woods v. Donald, 135 S. Ct. 1372, 1376 (2015) (citing White v. Woodall, 572 U.S. 415, 419 (2014)). However, Footnote 26 is not the only place in Duren where the Court makes clear that a showing of discriminatory intent is not required. In the body of the opinion, the Court enumerated the three elements that a prisoner must establish to prove a constitutional violation, thereby setting the outer parameters of a fair-cross-section analysis, and it simply did not include

8 discriminatory intent as one of those elements.2 Therefore, requiring a prisoner to show discriminatory intent imposes a more stringent standard than the one articulated by the Supreme Court. Though states may provide broader constitutional protections than required by federal law, they “may not impose . . . greater restrictions as a matter of federal constitutional law when [the Supreme] Court specifically refrains from imposing them.” Oregon v. Hass, 420 U.S. 714, 719 (1975) (emphasis omitted)). The state court did not address the three factors identified in the Duren test, but instead rested its decision exclusively on Howell’s failure to identify a discriminatory purpose. By requiring proof of this additional element, the Superior Court imposed greater restrictions on Howell than those required by the Supreme Court, contrary to and in an unreasonable application of clearly established federal law. B. Because the Superior Court’s decision contradicts federal law, this Court must review Howell’s claim de novo. To establish a fair-cross-section violation, Howell must prove that, at the time of his trial, (1) blacks were a “‘distinctive’ group in the community”; (2) “representation of [blacks] in

2 Writing in dissent, Justice Rehnquist criticized the majority for imposing the very distinction between Equal Protection Clause cases and Sixth Amendment cases that the Superior Court ignores. Duren, 439 U.S. at 371 (Rehnquist, J., dissenting) (emphasizing that “[t]he difference [between equal protection and Sixth Amendment cases] apparently lies in the fact, among others, that under equal protection analysis prima facie challenges are rebuttable by proof of absence of intent to discriminate, while under Sixth Amendment analysis intent is irrelevant”).

9 venires from which juries [were] selected [was] not fair and reasonable in relation to the number of such persons in the community”; and (3) “this underrepresentation [was] due to systematic exclusion of [blacks] in the jury selection process.” Duren, 439 U.S. at 364. 1. Distinctive Group Blacks are “unquestionably a constitutionally cognizable group.” Ramseur v. Beyer, 983 F.2d 1215, 1230 (3d Cir. 1992) (en banc). See also United States v. Weaver, 267 F.3d 231, 239 (3d Cir. 2001) (finding that blacks are “sufficiently numerous and distinct from others in the population” to satisfy the first prong of the Duren test (citing Castaneda v. Partida, 430 U.S. 482, 495 (1977))). 2. Unfair and Unreasonable Representation Howell’s claim that blacks were unfairly and unreasonably represented in jury venires “must be supported by statistical evidence,” beginning with the percentage of blacks in the community at the time of his trial. Weaver, 267 F.3d at 240 (citing Duren, 439 U.S. at 364). Relying on the 2000 Census, Howell has demonstrated that 10.7% of the adult population in Allegheny County identified as black. See Duren, 439 U.S. at 365 (accepting census data as “prima facie evidence of population characteristics”). This population percentage must then be compared to the percentage of blacks included in the jury venire to determine whether representation was proportionately fair and reasonable. Id. at 364-67. i. Reliability of the Data Howell relies on the 2001 study conducted by Gentile Meinert & Associates for his claim that blacks made up 4.87% of jury pools. However, there is no evidence regarding how many people received jury summonses, how many people

10 appeared for jury selection (versus the number of individuals who received surveys), or how many people failed to fill out the survey. Without this information, Howell’s statistical data is not sufficiently reliable to support a finding of unfair and unreasonable representation.3 See Weaver, 267 F.3d at 243-44. In Weaver, this Court found that a prisoner’s figures were too weak to support his claims where the statistician based his conclusions only on completed and returned questionnaires without accounting for unanswered questionnaires. Id. The Court highlighted that, to support an allegation of under-representation, the statistician was required to perform one of three analyses: (1) analyze the race of every person in the jury pool; (2) perform a sampling of the jury pool

3 Under AEDPA, the state court’s implicit and explicit factual findings are presumed correct “if supported by the record.” Taylor v. Horn, 504 F.3d 416, 433 (3d Cir. 2007); see also 28 U.S.C. § 2254(e)(1). Even if the Superior Court had implicitly made a credibility determination regarding Dr. Karns’ testimony—which it did not, compare Campbell v. Vaughn, 209 F.3d 280, 285 (3d Cir. 2000) (finding implicit credibility determination where Superior Court relied on the contested testimony to conclude that defendant did not demonstrate ineffective assistance of counsel), with App. at 252 (noting that Howell “relies on the testimony of John F. Karns, Ph.D.,” but then reaching its legal determination without any reference to or reliance upon Dr. Karns’ testimony)—that determination would be undermined by the record for the reasons we explain.

11 and then calculate the standard deviation4; or (3) account for the statistical impact of the unreturned questionnaires. Id. at 244. Because he did not provide any of these analyses, this Court concluded that the statistical evidence was “too weak to support a finding of representation that is unfair and unreasonable.”5 Id. Howell’s statistical data suffers from the same weaknesses we identified in Weaver. As in Weaver, Dr. Karns

4 “Standard deviation” is often confused with the similar, but distinct, calculation of “standard error.” See Douglas G. Altman & J. Martin Bland, Statistics Note, Standard Deviation and Standard Errors, 331 Brit. Med. J. 903 (2005). As called for in Weaver, reliable data requires a standard deviation calculation if the entire population is not accounted for, which “indicates how accurately the mean represents sample data.” Dong Kyu Lee et al., Standard Deviation and Standard Error of the Mean, 68 Korean J. Anesthesiology 220 (2015); see also Weaver, 267 F.3d at 238 n.6 (requiring calculation of the standard deviation “because it establishes the probability that a sample taken from the jury wheel accurately reflects the composition of the entire wheel”). 5 The Court also noted that discrepancies in the statistician’s testimony, wherein he consistently claimed to have examined the entire master wheel even though he did not account for unreturned surveys, “further undermine[d] the strength of the evidence.” Weaver, 267 F.3d at 243-44.

12 did not analyze the racial makeup of the entire jury venire.6 Though approximately 4500 individuals were given surveys over a six-month period, Dr. Karns’ analysis did not take the unanswered surveys into consideration, which significantly weakens the reliability and influence of the statistical data. Id. at 244. As Dr. Karns acknowledged, if a higher percentage of blacks failed to answer the survey than whites, the results of the survey would be “skewed.” App. at 131. However, Dr. Karns does not know how many surveys omitted responses to certain questions or went unanswered entirely, let alone the race of the individuals who chose not to answer them. Because of this missing data, it is not possible to now calculate the standard deviation or account for the significance of unanswered surveys, as we require. Howell claims that Dr. Karns’ data does satisfy Weaver because he conducted a validity analysis known as the “Z- statistic,” which Howell claims is “akin to standard deviation,” and concluded that the chances of his conclusion that blacks were under-represented being incorrect “are about four in 10,000.” Reply Br. at 13 (quoting App. at 112). However, the purpose of the “Z-statistic” is simply to determine the “risk of being wrong” about a hypothesis. App. at 112. Here, Dr. Karns’ starting hypothesis was “that there are too few African- Americans” in jury pools. Id. However, Dr. Karns did not

6 In addition to acknowledging that he had “no idea” whether every potential juror filled out the survey, App. at 117—and it would be illogical to believe that each person did—Dr. Karns also testified that jurors who were originally assembled in civil court assignment rooms but were later brought to criminal court were not surveyed. Therefore, we can conclude without speculation that Dr. Karns’ analysis failed to account for every member of the venire.

13 provide any analysis to explain how a low likelihood of this hypothesis being incorrect sufficiently demonstrates that his statistical representations are reliable, particularly in light of the unaccounted for, unanswered surveys. For instance, it could certainly be true that blacks appear on jury pools less often than we would statistically expect, but that the degree of under-representation does not rise to the level of a constitutional violation. Dr. Karns’ Z-statistic analysis regarding the accuracy of his general hypothesis cannot substitute a standard deviation calculation, which is an inquiry into the reliability of the statistics he presented and is required by our precedent. Because Howell’s statistical data fails to account for the entire jury venire using one of the statistical methodologies approved by this Court, it is “too weak to support a finding of representation that is unfair and unreasonable.” Weaver, 267 F.3d at 244. ii. Significance of the Data Even if Howell had provided reliable data, courts around the nation, including our own, have found that representation was not unfair or unreasonable with disparity levels greater than or similar to those presented here. To determine the significance of the statistical evidence, we must compare the population percentage (10.7%) with the jury venire percentage (4.87%). This Court has relied on two methods of statistical analysis to determine the significance of

14 the disparity between the percentages: absolute disparity7 and comparative disparity.8 Weaver, 267 F.3d at 241; Ramseur, 983 F.2d at 1233-35. The absolute disparity in this case, 5.83%, is lower than or similar to absolute disparities in other cases where courts have found no constitutional violation, and in fact, numerous courts have noted that an absolute disparity below 10% generally will not reflect unfair and unreasonable representation. See United States v. Shinault, 147 F.3d 1266, 1273 (10th Cir. 1998) (noting that courts of appeals “generally are reluctant to find [unfair and unreasonable representation] when the absolute disparities are less than 10%”); see also,

7 Absolute disparity reflects the difference in the percentage of, in this case, blacks in the general voting-age population and in the jury venire: 10.7% (population percentage) - 4.87% (venire percentage) = 5.83% (absolute disparity). This absolute disparity reflects that, in a jury pool of one hundred people, approximately six fewer black people would be in the pool than statistically expected. 8 Comparative disparity “measures the decreased likelihood that members of an underrepresented group will be called for jury service” relative to what would be expected given the percentage of the general population that group comprises. United States v. Shinault, 147 F.3d 1266, 1272 (10th Cir. 1998) (emphasis in original) (cited by Weaver, 267 F.3d at 241-42). This is calculated by dividing the absolute disparity by the population percentage: 5.83% (absolute disparity) ÷ 10.7% (population percentage) = 54.49% (comparative disparity). This comparative disparity reflects that, at the time of Howell’s trial, blacks were 54.49% less likely to be on venires than if the representation was directly proportional to their population in the County.

15 e.g., Thomas v. Borg, 159 F.3d 1147, 1151 (9th Cir. 1998) (5% absolute disparity insufficient even though no blacks were on jury panel); United States v. Gault, 141 F.3d 1399, 1402-03 (10th Cir. 1998) (3.19%, 5.74%, and 7.0% absolute disparities insufficient); United States v. Pion, 25 F.3d 18, 23 (1st Cir. 1994) (3.4% absolute disparity insufficient); Ramseur, 983 F.2d at 1232 (absolute disparity of 14.1% “borderline”); United States v. Suttiswad, 696 F.2d 645, 649 (9th Cir. 1982) (2.8%, 4.7%, and 7.7.% absolute disparities insufficient). Likewise, courts have found that comparative disparities similar to the comparative disparity in this case, 54.49%, were insufficient to demonstrate unfair and unreasonable representation. See, e.g., United States v. Chanthadara, 230 F.3d 1237, 1257 (10th Cir. 2000) (finding comparative disparity of 40.89% insufficient where the distinctive group represented 7.9% of the population); United States v. Clifford, 640 F.2d 150, 155-56 (8th Cir. 1981) (finding comparative disparity of 46% insufficient where the group represented 15.6% of the population). But see LaRoche v. Perrin, 718 F.2d 500, 502-03 (1st Cir. 1983) (holding that a prima facie challenge was established where the comparative disparity was 68.22% and the group comprised 38.4% of the population), overruled on other grounds by Barber v. Ponte, 772 F.2d 982 (1st Cir. 1985). When compared to factually similar cases, the absolute and comparative disparities reflected in this case do not make a prima facie showing of unconstitutional under- representation. 3. Systematic Exclusion If Howell’s claims were supported by reliable statistical evidence, to prove a cross-section violation, Howell would need to show that the under-representation of blacks in jury

16 pools is “due to systematic exclusion in the jury selection process.” Weaver, 267 F.3d at 244 (citing Duren, 439 U.S. at 366). In Duren, the Supreme Court found systematic exclusion where a state law permitted women to exclude themselves from jury selection simply because of their gender. 439 U.S. at 367. Unlike in Duren, where the system that caused the under- representation—a state statute—was readily apparent, there is no identifiable cause for the under-representation of blacks in jury venires in Allegheny County. Therefore, to demonstrate “systematic exclusion,” Howell must show “a large discrepancy over time such that the system must be said to bring about the underrepresentation.” Weaver, 267 F.3d at 244. We consider the nature of the system, length of time studied, and “efforts at reform to increase the representativeness of jury lists” in determining whether the jury selection system caused the under-representation. Ramseur, 983 F.2d at 1234-35. i. Nature of the System A selection process that is facially neutral is unlikely to demonstrate systematic exclusion. See Ramseur, 983 F.2d at 1235. In Ramseur, we concluded that the selection process was facially neutral because the pool of jurors (the “Master List”) was composed of names from both the voter registration and Department of Motor Vehicles lists, and, therefore, did not preference any particular age, gender, or race. Id. Likewise, at the time of Howell’s trial, the Master List consisted of names from Allegheny County’s list of registered voters and the Pennsylvania Department of Transportation’s driving records. Howell does not contest the propriety of Allegheny County’s method for compiling its Master List, and these parallels demonstrate that the nature of the system was facially neutral.

17 ii. Length of Time Studied Even assuming that Howell’s data was based on a reliable study, that study must have demonstrated ongoing discrimination over a sufficient period of time. In Ramseur, this Court held that a study conducted over the course of two years was not sufficient to show a history of abuse that would reflect a systematic exclusion. 983 F.2d at 1235. Howell seeks to distinguish the six-month study in this case from Ramseur by noting that, in Duren¸ the underlying study lasted for only eight months.9 Howell cannot distinguish his case from Ramseur by relying on the eight-month study in Duren because the problematic system there—a gender-based exemption statute—was readily identifiable and undisputed. Duren, 439 U.S. at 367. Additionally, unlike here, where the data reflects an amalgamation of the racial makeup of jury pools over the six-month period, Duren undisputedly demonstrated “that a large discrepancy occurred not just occasionally but in every weekly venire for a period of nearly a year.” Id. at 366. The Supreme Court emphasized that this repeated, perpetual under- representation “manifestly indicate[d] that the cause of the under-representation was systematic.” Id. Howell’s evidence is not similarly specific and does not support a conclusion that the under-representation was occurring in every, or even nearly

9 On appeal, Howell also points to media reports and studies regarding racial under-representation that began in 2002; however, these studies were not part of the record before the state court, and we cannot consider them. See S.H. ex rel. Durrell v. Lower Merion Sch. Dist., 729 F.3d 248, 267 n.27 (3d Cir. 2013) (refusing to consider evidence offered for the first time on appeal).

18 every venire for a substantial period of time. iii. Efforts to Reform Where the government is engaged in on-going efforts to improve the representativeness of jury lists, it is less likely that the data reflects that under-representation is due to a systematic exclusion in the jury process. Ramseur, 983 F.2d at 1235. We presume that the process is legitimate where the government’s efforts seem likely to create a representative jury, even if the statistical evidence demonstrates that the pool is “not representative enough.” Id. At the time of Howell’s trial, Allegheny County was unable to say whether there was a representation problem with its Master List because its records did not reflect the races of potential jurors. Around 2002, to remedy the risk of under- representation, the Court Administration Office revised its eligibility questionnaire to include questions regarding race, age, and gender so that it could better understand whether a particular group was over-represented or under-represented. Allegheny County additionally implemented procedures to follow up on unreturned questionnaires, ensure that the Master List reflects up-to-date addresses, and encourage individuals to respond to jury summonses. According to the Court Administration Office, each of these actions was implemented to better ensure proportionate representation. These laudable remedial actions warrant “some presumption of [the jury system’s] legitimacy,” Ramseur, 983 F.2d at 1235, and reflect that Allegheny County’s processes were not systematically exclusive. IV. Though the Pennsylvania Superior Court misapplied the Supreme Court’s precedent in denying Howell’s Sixth Amendment claim, on de novo review, we find that Howell

19 failed to show that Allegheny County’s jury selection processes systematically excluded black jurors. We will therefore affirm the District Court’s denial of habeas relief.

20 PORTER, Circuit Judge, concurring.

I join the majority in holding that Joseph Howell failed to satisfy the second and third requirements of Duren v. Missouri, 439 U.S. 357 (1979). But I reach that conclusion slightly differently. On Duren’s second prong, I would avoid the soundness-of-the-statistics debate for a simple reason: even assuming arguendo that Howell’s statistics are methodologically sound, the disparity figures are within the range that we have held constitutionally permissible. So I would hold that Howell fails Duren’s second requirement on that basis. On Duren’s third requirement, I agree with the majority’s analysis. But I supplement it to underscore that Allegheny County’s jury-selection system goes above and beyond what is constitutionally required, so there cannot be systematic exclusion.

To satisfy Duren’s second requirement, a defendant must show that “the representation of [an underrepresented distinctive] group in jury venires is not ‘fair and reasonable’ in relation to the number of such persons in the community.” United States v. Weaver, 267 F.3d 231, 237 (3d Cir. 2001) (citing Duren, 439 U.S. at 364). As the majority observes, two statistical measurements drive this analysis: absolute disparity and comparative disparity. We consider both of these disparity measures, which makes us something of an outlier. See Nancy Gertner, et al., The Law of Juries § 2.11 (10th ed. 2018) (noting that while “[t]he Supreme Court has not mandated the use of one approach over another,” in practice, “[m]ost [courts] have rejected comparative disparity analysis”).

Howell’s statistics show an absolute disparity of 5.83%, which is easily within the range typically found constitutionally permissible. As the leading treatise

1 summarizes, “[m]any courts have adopted a threshold of 10% absolute disparity.” Gertner, § 2.12. We have followed this trend, marking the threshold a smidge higher. See Ramseur v. Beyer, 983 F.2d 1215, 1232 & n.18 (3d Cir. 1992) (“Courts addressing the question of whether a given absolute disparity constitutes ‘substantial underrepresentation’ have held that absolute disparities between 2.0% and 11.5% do not constitute substantial underrepresentation.” (quoting Castaneda v. Partida, 430 U.S. 482, 494 (1977))). So the absolute disparity of 5.8% in this case is constitutionally permissible under authorities from this and other courts.

This means that Howell must rely on comparative disparity to satisfy Duren’s second prong. This is a much closer question. Under our precedents, the comparative disparity of 54.5% shown here is troubling. Ramseur, 983 F.2d at 1232 (describing “a comparative disparity of about 40%” as “borderline” but ultimately rejecting prima facie case); see also Weaver, 267 F.3d at 243 (describing comparative disparity figures of 40.01% for blacks and 72.98% for Hispanics as “quite high,” but qualified that the figures were of limited value because both groups formed “a small percentage of the population”). But we have never held that a high comparative disparity is itself sufficient to satisfy Duren’s second prong. And indeed, other courts have rejected fair-cross-section challenges involving comparative disparities higher than (or similar to) the one here.1 So the comparative-disparity figure

1 See, e.g., United States v. Shinault, 147 F.3d 1266, 1273 (10th Cir. 1998) (permitting comparative disparities of “48%, 50%, and almost 60%”); United States v. Chanthadara, 230 F.3d 1237, 1257 (10th Cir. 2000) (permitting “a comparative disparity of 58.39%”); United States v. Sanchez, 156 F.3d 875,

2 in this case—while high—is not enough to satisfy Duren’s second prong.

Turning to Duren’s third requirement, Howell must show “the underrepresentation is caused by the ‘systematic exclusion of the group in the jury selection process.’” Weaver, 267 F.3d at 237 (quoting Duren, 439 U.S. at 364). On this point, I am puzzled by the dissent’s insistence that the County’s system is constitutionally deficient.

The County’s two-track method of selecting jurors is structurally sound. It first draws names from voter-registration lists. It then supplements this by pulling additional names from motor-vehicle records. If anything, the County’s system goes above and beyond what is required, as courts have consistently held that using voter-registration lists alone is sufficient.2 “Not

879 & n.4 (8th Cir. 1998) (acknowledging a comparative disparity of 58.3%, but declining to address statistics at all to “simply hold that when jury pools are selected from voter registration lists, statistics alone cannot prove a Sixth Amendment violation”); Hafen, 726 F.2d at 23–24 (permitting comparative disparity of 54.2%); United States v. Sanchez- Lopez, 879 F.2d 541, 548–49 (9th Cir. 1989) (permitting comparative disparity of 52.9%); United States v. Orange, 447 F.3d 792, 798–99 (10th Cir. 2006) (permitting comparative disparity of 51.22%). 2 United States v. Guzman, 468 F.2d 1245, 1247–49 (2d Cir. 1972) (approving the use of voter-registration lists as the sole source of names for jury selection); United States v. Odeneal, 517 F.3d 406, 412 (6th Cir. 2008) (approving jury administrator’s use of voter-registration lists, noting these “are the presumptive statutory source for potential jurors”) (citing

3 only has the use of the voter registration lists been uniformly approved by the Court[s] of Appeals as the basic source for the jury selection process … Congress specifically approved the use of such lists even though it was recognized that persons who chose not to register would be excluded from the jury selection process.” United States v. Cecil, 836 F.2d 1431, 1448 (4th Cir. 1988) (citing 28 U.S.C. § 1863(b)(2)). In fact, the County’s two-track system here is strikingly similar to the one we upheld in Ramseur. 983 F.2d at 1233 (noting that the “mechanism used to create the source lists was facially neutral with respect to race,” as the New Jersey county in question “utilized voter registration and Department of Motor Vehicle lists to create its jury venire”).

Unsurprisingly, then, the dissent cites no case in which a hybrid system like this one—i.e., voter-registration lists supplemented with motor-vehicle records—has been held to systematically exclude a distinctive group. In dicta, we have speculated “that if the use of voter registration lists over time did have the effect of sizeably underrepresenting a particular class or group on the jury venire, then under some circumstances, this could constitute a violation of a defendant’s fair cross-section rights under” the Sixth Amendment. Weaver, 267 F.3d at 244–45 (internal quotation marks and citation omitted). But that theoretical possibility was not the reality in Weaver, as “nothing in the record” showed persistent systematic exclusion of minority jurors. Id. at 245. And

28 U.S.C. § 1863(b)); United States v. Greatwalker, 356 F.3d 908, 911 (8th Cir. 2004) (finding no systematic exclusion from jury selection plan that draws its pools of prospective jurors randomly from lists of persons who voted in the last presidential election).

4 whatever the merits of that theoretical possibility, we have never invoked it to hold that a hybrid system like this one systematically excluded a distinctive group. Given that Congress has made voter-registration lists the presumptive source for selecting jurors, such a holding could imperil juror- selection methods across many jurisdictions.

In support of systematic exclusion, Howell argues that the County’s problems with “non-representative jury venires were widely known well before” Howell’s trial, largely because the County and some academics studied it. Appellant’s Br. 36–39. This is weak tea. The fact that the County studied this issue does not show that the County knew its selection system was constitutionally unsound; rather, it may simply show that the County was responsibly trying to determine the system’s soundness or seeking to improve (already constitutionally sufficient) representation. In Ramseur, we viewed a New Jersey county’s efforts to diversify jury venires just this way, approvingly noting the county’s “efforts at reform to increase the representativeness of jury lists.” 983 F.2d at 1235. Howell’s inferences, by contrast, would perversely punish the County for its salutary reform efforts.

In sum, if the County used only voter-registration lists to assemble the jury venire, it would be employing a method widely upheld as constitutional by the courts of appeals and statutorily prescribed by the Jury Selection and Service Act. 28 U.S.C. §§ 1861–78. By supplementing this method with motor-vehicle records, the County goes beyond this widely approved method to mirror the system upheld in Ramseur. Howell has not suggested how the County could improve upon this system and I see no constitutional requirement for it to do so.

5 RESTREPO, Circuit Judge, concurring in part and dissenting in part.

I join the majority opinion only with respect to Part III.A, in which the majority holds that we are not required to accord deference under the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”) to the legal conclusions of the Pennsylvania Superior Court because that court’s decision was contrary to, or involved an unreasonable application of, clearly established federal law. I respectfully dissent from the remainder of the majority opinion because, in my view, Howell has established a prima facie violation of his Sixth Amendment right to have his petit jury drawn from a fair cross-section of the community, and I would reach the merits of his fair-cross- section claim because the Commonwealth has presented no evidence to rebut Howell’s statistical analysis or the qualifications of his expert witness. The majority, however, lends undue credence to the Commonwealth’s speculative attack on the reliability of Howell’s statistics and, in the process, sets forth a new standard of statistical purity that will foreclose nearly all fair-cross-section claims. And with respect to the merits of Howell’s fair-cross-section claim, the majority and concurring opinions interpret the case law in a way that deprives the Sixth Amendment of any power to provide a remedy in cases where a distinctive group that constitutes less than 10% (or, for the concurrence, 11.5%) of the population is systematically excluded from serving on venires, even if the entire group is completely excluded from venire service. Such an interpretation simply cannot be an accurate statement of the law.

1 I.

Howell presented evidence that black persons constituted 10.7% of the jury-service-eligible population of Allegheny County in the early 2000s but merely 4.87% of persons serving on venires during the same period. Thus, according to Howell’s evidence, black persons in Allegheny County were underrepresented on venires by approximately 54.49%. Put another way, it appears that over half of Allegheny County’s black jury-service-eligible population—a significant population of nearly 110,000 people—was excluded from serving on venires.

Rather than discussing these troubling statistics at length, the majority simply attacks their reliability. In so doing, the majority misapplies our precedent in United States v. Weaver, 267 F.3d 231 (3d Cir. 2001), and, as a result, sets a new bar for statistical reliability that almost no litigant in a fair- cross-section case will be able to satisfy.1

1 Independently, the Court also may lack authority under AEDPA to probe into the reliability of Howell’s statistics in the first place. Pursuant to AEDPA, in a section 2254 proceeding such as this one, “a determination of a factual issue made by a State court shall be presumed to be correct.” 28 U.S.C. § 2254(e)(1). Both implicit and explicit factual findings are presumed to be correct under section 2254(e)(1). Taylor v. Horn, 504 F.3d 416, 433 (3d Cir. 2007). Two of the three judges on the panel of the Superior Court appear to have reached their decisions by taking Howell’s statistical evidence at face value, which, in my opinion, may constitute an implicit factual finding that is entitled to the “presumption of correctness” under section 2254(e)(1). See App. 258.

2 The majority reads Weaver as requiring all litigants asserting fair-cross-section claims to either (1) produce documentary evidence that they conducted a complete census of the races of every single individual in the relevant jury pool (e.g., every person on the “master wheel” or venire), or (2) perform sampling of the jury pool “and then calculate the standard deviation,” or (3) “account for the statistical impact” of persons in the jury pool who were not surveyed or studied. 267 F.3d at 244. This reading of Weaver disregards the specific context of that case. In Weaver, the demographer who provided expert testimony regarding the racial makeup of the “master wheel” in the Erie Division of the Western District of Pennsylvania purported to have studied all persons on the “master wheel,” on which 5,877 persons were listed. See id. at 243. Our Court determined, however, that the demographer “based his testimony on the returned questionnaires,” of which there were only 4,753. Id. Thus, in Weaver, concrete evidence—figures that demonstrated with specificity that 1,124 persons, or over 19%, of the relevant jury pool were not included in the study—effectively impeached the demographer’s testimony that he had studied all persons in the jury pool. Consequently, because the demographer did not— either quantitatively or qualitatively—account for the glaring discrepancy in his testimony, our confidence in the reliability of his statistics was undermined.

Placed in context, Weaver stands for the proposition that “the strength of [a litigant’s statistical] evidence” is “undermined” when (1) the state produces concrete evidence that the petitioner’s expert did not study all persons in the relevant jury pool and (2) the expert neither (A) “perform[ed] sampling” of the jury pool “and then calculate[d] the standard

3 deviation” nor (B) “account[ed] for the statistical impact of” unstudied or uncounted persons in the jury pool. Id. at 244.

Here, there is no such concrete evidence that Howell’s expert failed to study all persons on the venires during the six- month study period—there is only speculation. Despite its failure to substantively challenge the reliability of Howell’s statistics or the qualifications of Howell’s expert in any of the state-court proceedings below, the Commonwealth, in its brief, now argues that the Court should disregard Howell’s statistical evidence solely because his expert, Dr. John F. Karns, Ph.D., “did not know if every individual [in the studied venires] complied with the request to fill out the questionnaire[s].” Appellee’s Br. 15. The Commonwealth presents no evidence regarding the number of veniremembers who allegedly did not return the questionnaires; it merely speculates that there could have been veniremembers who did not return the questionnaires.

For the majority, mere speculation of this nature is sufficient to defeat Howell’s Sixth Amendment fair-cross- section claim. This holding—that the state can defeat a fair- cross-section claim simply by speculating, with no evidentiary support, that a habeas petitioner’s statistics may be flawed— transforms the modest holding in Weaver regarding statistical reliability into a holding that dramatically heightens the burden of proof in fair-cross-section cases. In effect, the majority holds that, to state a Sixth Amendment fair-cross-section claim, a litigant must produce unassailable proof that she conducted a complete census of every single member of the relevant jury pool; if the state simply speculates that certain members of the jury pool may have been excluded from the study, and even if the state provides zero evidence to that

4 effect, the litigant’s fair-cross-section claim fails unless certain limited conditions are met.

The majority also takes a severely constrained view with respect to what evidence can satisfy such limited conditions and requires Howell to produce evidence that is wholly irrelevant to its inquiry into the reliability of his statistics. Relying on its reading of Weaver, the majority holds that because Howell’s statistical analysis is fundamentally undermined by the Commonwealth’s speculation regarding the potential existence of unstudied veniremembers,2 Howell’s claim may only survive if he either (1) “calculate[s] the standard deviation” or (2) “account[s] for the statistical impact of . . . unreturned questionnaires.” Howell has produced evidence that satisfies both of these conditions, even assuming that both conditions are relevant. Regarding the “significance of unanswered surveys,” the only concrete evidence in the record that indicates that certain veniremembers were omitted

2 As an ancillary matter, the majority also holds that Howell’s statistical evidence is undermined by the fact that “there is no evidence regarding how many people received jury summonses.” It is unclear how information with respect to “how many people received jury summonses” is relevant to Howell’s claim because his claim is based on the composition of the venires—the persons who actually appeared for jury service—in Allegheny County, a type of claim that has long been recognized as cognizable by the Supreme Court. See, e.g., Taylor v. Louisiana, 419 U.S. 522, 538 (1975) (“[T]he jury wheels, pools of names, panels, or venires from which juries are drawn must not systematically exclude distinctive groups in the community and thereby fail to be reasonably representative thereof.” (emphasis added)).

5 from the study is that “a very small number” of “surveys contain[ed] incomplete information.” App. 118. Dr. Karns explicitly testified as to the statistical impact of these incomplete surveys on his results: the number of such surveys was “so small that it [did] not change [his] opinion.” Id. at 128. Thus, Howell has accounted for the only concrete evidence in the record that his statistical analysis may be based on less than complete information, and, therefore, Howell has satisfied one of the majority’s requirements.

Regarding the majority’s requirement that Howell calculate the standard deviation, it is not clear to me how calculation of the standard deviation relates to the question that the majority seeks to answer: How do (potentially) unaccounted-for veniremembers affect the reliability of Howell’s statistical analysis? “[S]tandard deviation is a measure of [the] variability . . . of the population from which [a] sample was drawn.”3 In other words, standard deviation is an expression of “how widely scattered some measurements [of a population] are.”4 For example, students who score a 141 on the LSAT have scores that are one standard deviation from

3 Douglas G. Altman & J. Martin Bland, Statistics Note, Standard Deviations and Standard Errors, 331 Brit. Med. J. 903 (2005), https://www.ncbi.nlm.nih.gov/pmc/ articles/PMC1255808/pdf/bmj33100903.pdf (emphasis added). 4 Id.

6 the mean score of 151.5 But the fact that one standard deviation is equivalent to approximately 10 points in the context of the distribution of LSAT scores tells us nothing about the statistical reliability of the analysis conducted by the Law School Admission Council—it only tells us how the scores are distributed on a curve. It appears to me that the majority actually desires a calculation of the “standard error,” which “indicates the uncertainty around the estimate of the mean” due to, among other things, sampling errors.6 “The terms ‘standard error’ and ‘standard deviation’ are often confused.”7 The former concept, standard error, concerns the reliability of Howell’s statistics, which statistics indicate that over the course of the study period, a mean of 4.87 black persons served on every venire of 100 persons; standard error would tell us how confident we should be that the mean of 4.87 is an accurate figure. In requiring that Howell instead calculate the standard deviation, the majority perpetuates an error of terminology first committed by our Court in Weaver. See 267 F.3d at 244 (“In order to support Weaver’s allegation of underrepresentation on the master wheel, [his expert] would have had to . . . calculate the standard deviation . . . .”). Thus,

5 See Memorandum from Lisa Anthony, Senior Research Assoc., Law Sch. Admission Council, to LSAT Score Recipients 2 (June 20, 2017), https://www.lsac.org/sites/default/files/legacy/docs/default- source/data-%28lsac-resources%29-docs/lsat-score- distribution.pdf. 6 Altman & Bland, supra note 3. 7 Id.

7 the majority requires Howell to produce evidence that is not at all relevant to probing the reliability of his statistics.8

8 If, however, the majority truly desires a calculation of the standard deviation—which is irrelevant for the reasons stated above—Howell has produced equivalent statistical evidence. Dr. Karns used a “difference-of- proportion test” by calculating a “Z-statistic,” App. 112, and then calculating what social scientists refer to as a “P value,” which is a “statistical summary of the compatibility between the observed data and what we would predict or expect to see if we knew the entire statistical model.” Sander Greenland et al., Statistics Tests, P Values, Confidence Intervals, and Power: A Guide to Misinterpretations, 31 Eur. J. Epidemiology 337, 339 (2016). Put differently, a P value “can be viewed as a continuous measure of the compatibility between the data and the entire model used to compute it, ranging from 0 for complete incompatibility to 1 for perfect compatibility.” Id. Similar to the way that standard deviation indicates the variance within a population, a P value indicates the variance between observed data and the data that we would expect to observe. Here, for instance, we would expect that the percentage of black persons serving on venires in Allegheny County would mirror the black jury-service-eligible population of Allegheny County as a whole (10.7%). As Dr. Karns observed, however, black persons constituted merely 4.87% of persons serving on venires. That observed data (4.87%) varies widely from the expected data (10.7%), resulting in a P value of .0004 according to Dr. Karns, which closely nears complete incompatibility. See App. 112 (characterizing the “chances of being wrong in stating that there are too few African[ ]Americans” as “about four in 10,000”). Statisticians often

8 Further, standing alone, the sample size of the study upon which Howell relies indicates that Howell’s statistics are reliable. Approximately 4,500 persons were surveyed in connection with the study. Unrebutted expert testimony in this case establishes that a “sample of 4[,]500 is relatively large.” App. 119. Because the sample in this case was so large, the standard error necessarily is small because “[t]he standard error falls as the sample size increases, as the extent of variation is reduced.”9 By questioning the reliability of the statistics resulting from such a large sample size and by emphasizing the alleged importance of surveying every single member of venires without exception, the majority undermines the very concept of sampling in Sixth Amendment challenges.

In sum, the majority opinion sets forth a new standard of statistical purity that appears to be unattainable for nearly all litigants—and particularly for habeas petitioners—in fair- cross-section cases. Litigants are required to present statistical evidence to support fair-cross-section claims. See Duren v. Missouri, 439 U.S. 357, 364 (1979). If the state can fundamentally undermine a litigant’s statistical analysis with mere speculation that her statistics are unreliable, nearly all

characterize P values in terms of “the probability that chance alone produced the observed association.” Greenland et al., supra, at 340. Thus, if the majority desires statistical evidence regarding variance—which is what standard deviation expresses—Howell has provided such evidence to the Court in the form of a P value. 9 Altman & Bland, supra note 3.

9 force has been drained from the Sixth Amendment’s fair-cross- section requirement.

II.

Accepting the reliability of his statistical evidence, Howell, in my view, has satisfied both the second and third prongs of the test espoused by the Supreme Court in Duren v. Missouri, 439 U.S. at 364;10 namely, he has demonstrated that (A) “the representation” of black persons “in venires from which juries are selected is not fair and reasonable in relation to the number of such persons in the community” and (B) “this underrepresentation is due to the systematic exclusion of this group in the jury-selection process.”

A.

Howell has demonstrated that black persons in Allegheny County were underrepresented on venires by approximately 54.49% in the early 2000s. This rate of underrepresentation simply cannot be “fair and reasonable” under Duren.

“[N]either Duren nor any other decision of th[e Supreme] Court specifies the method or test courts must use to measure the representation of distinctive groups in jury pools.”

10 As the majority recognizes, Howell undoubtedly has satisfied Duren’s first prong, which requires him to demonstrate that black persons are “a ‘distinctive’ group in the community.” Duren, 439 U.S. at 364; see also Ramseur v. Beyer, 983 F.2d 1215, 1230 (3d Cir. 1992) (en banc) (holding that black persons are “unquestionably a constitutionally cognizable group”).

10 Berghuis v. Smith, 559 U.S. 314, 329 (2010). Our Court previously has utilized “absolute disparity” and “comparative disparity” to analyze the merits of fair-cross-section claims. Weaver, 267 F.3d at 241 & n.11. “Absolute disparity” is the “difference between [(x)] the percentage of a certain population group eligible for jury duty and [(y)] the percentage of that group who actually appear in the venire.” Ramseur v. Beyer, 983 F.2d 1215, 1231 (3d Cir. 1992) (en banc). “Comparative disparity is calculated by dividing [(x)] the absolute disparity by [(y)] the population figure for a population group.” Id. Although “both methods have been criticized,” Weaver, 267 F.3d at 242, we have held that “figures from both methods inform the degree of underrepresentation,” and we “examine and consider the results of both in order to obtain the most accurate picture possible,” id. at 243.

The comparative disparity in this case is 54.49%, while the absolute disparity in this case is 5.83%. The Commonwealth argues that analysis of the absolute disparity is the “starting place” when considering a fair-cross-section challenge and that, given the absolute-disparity figure in this case, it also should be the ending place for Howell’s fair-cross- section claim. Appellee’s Br. 19. Relying on dicta in our decision in Ramseur v. Beyer, 983 F.2d 1215, the Commonwealth argues that “[a]bsolute disparities between 2.0% and 11.5% have not constituted substantial underrepresentation” and that, “[t]herefore, under applicable precedent, an [a]bsolute [d]isparity of 5.83% is statistically insufficient to demonstrate a prima facie showing of a Sixth Amendment violation.” Appellee’s Br. 20 (emphasis omitted). This argument not only disregards our Court’s observation that “[o]ur precedent does not dictate that one method of statistical analysis should be used rather than another,” Weaver, 267 F.3d

11 at 241, but also misapprehends what the absolute-disparity figure captures. Viewed in isolation, an absolute-disparity figure lacks any meaning because the same absolute-disparity figure can imply drastically different levels of underrepresentation in two distinct populations.

For example, if, as the Commonwealth seems to suggest, an absolute disparity of over 11.5% is required for a litigant to state a Sixth Amendment fair-cross-section claim, Howell would never be able to state a fair-cross-section claim; the black jury-service-eligible population of Allegheny County is 10.7%, and thus the maximum absolute disparity in Howell’s case is 10.7%, which assumes the complete exclusion of black persons from service on venires (i.e., a comparative disparity of 100%). By contrast, in Philadelphia County, for example, which has a black population of approximately 43.4%, an absolute disparity of 11.5% would equate to underrepresentation of black persons on venires at a rate (and a comparative disparity) of 26.5%, raising much fewer constitutional concerns. It approaches absurdity to argue that the entire black population of Allegheny County could be excluded from serving on venires without violating the Constitution simply because a single metric—absolute disparity—is not high enough, without reference to any other factors.

But the majority and concurring opinions adopt precisely that argument. The majority holds that “an absolute disparity below 10% generally will not reflect unfair and unreasonable representation.” The concurrence takes this line of argument even further, framing an absolute disparity of 10% as a “threshold” matter and implying that this Court has set the “threshold” at the even higher figure of 11.5%. By definition,

12 the absolute disparity in a given case can only be as high as the percentage of the population that a distinctive group constitutes. If a litigant must present evidence of an absolute disparity of 10% (or, for the concurrence, 11.5%) as a “threshold” matter to state a fair-cross-section claim, then litigants, as a matter of law, cannot state fair-cross-section claims if the distinctive group that they allege was systematically excluded from serving on venires constitutes less than 10% (or 11.5%) of the population because, in such a case, even complete exclusion of such a group would not result in an absolute disparity of 10% (or 11.5%). In essence, the majority and concurring opinions hold that the Sixth Amendment provides no remedy for complete, systematic exclusion of distinctive groups in the community if those groups constitute less than 10% (or 11.5%) of the population.

Both the majority and concurring opinions also misunderstand the interaction between absolute disparity and comparative disparity. Analyzing the absolute disparity and comparative disparity in a case is not an either-or proposition: “figures from both methods inform the degree of underrepresentation.” Id. at 243 (emphasis added). We look at both figures because comparative disparity is a dependent variable—in fact, absolute disparity is the numerator in the formula used to calculate comparative disparity. In other words, we cannot even calculate the comparative disparity in a case without knowing the absolute disparity. Thus, the comparative disparity in a case, by necessity, implies a precise absolute disparity—every comparative disparity has a corresponding absolute disparity, and vice versa.

If, as the majority and concurring opinions hold, a litigant must present evidence of an absolute disparity of 10%

13 (or 11.5%) as a “threshold” matter to state a fair-cross-section claim, the opinions’ analyses of the comparative disparity in Howell’s case are merely perfunctory. As illustrated in the Appendix to this opinion, Howell would have to produce evidence of a comparative disparity of 93.46% or higher to satisfy a 10% absolute-disparity “threshold,” and Howell could never satisfy a 11.5% absolute disparity “threshold” because he would have to produce evidence of a comparative disparity in excess of 100%, which is impossible. If—as the majority and concurring opinions, by necessity, hold—the comparative disparity in Howell’s case must exceed these figures because absolute disparity is a “threshold” matter, any analysis in the majority and concurring opinions with respect to the sufficiency of Howell’s comparative disparity figure of 54.49% necessarily must be composed of empty words.

In my view, Howell’s statistics are sufficient to state a fair-cross-section claim. When analyzing this case, my reading of the case law compels me to start with the comparative disparity of 54.49%. This figure—which implies that over half of Allegheny County’s black jury-service-eligible population was excluded from serving on venires—should trouble everyone. Although this figure is well above the 40% figure that we called “borderline” in Ramseur, 983 F.2d at 1232, our analysis cannot stop there because we have recognized that comparative disparity may overstate the degree of underrepresentation in cases “where a small population is subjected to scrutiny,” Weaver, 267 F.3d at 242.

We must, then, look at the size of the population at issue—and, consequently, at the absolute disparity—to place the troubling 54.49% comparative disparity into context and determine whether it rises to the level of a Sixth Amendment

14 violation. See id. (“[T]he significance of the [comparative- disparity] figure is directly proportional to the size of the group relative to the general population . . . .”). For example, in Weaver, we noted that comparative disparities of 40.01% with respect to black persons and 72.98% with respect to Latino persons were “quite high,” but because the black and Latino jury-service-eligible populations constituted merely 3.07% and 0.97% of the total jury-service-eligible population, respectively, we held that these figures did not rise to an unconstitutional level of underrepresentation. Id. at 238, 243. In essence, because the populations at issue in Weaver were so small—resulting in absolute disparities of 1.23% for black persons and 0.71% for Latino persons—the net impact of the underrepresentation of these racial groups on venires was minimal, and therefore their degree of representation on venires was “fair and reasonable” under the Sixth Amendment. See id. at 243.

Here, we are not confronted with a small population group as in Weaver; rather, we are confronted with a group that constitutes over one-tenth—10.7%—of the relevant jury- service-eligible population. Given the significant size of that group—black persons—as a proportion of the total jury- service-eligible population, underrepresentation of black persons at a rate of 54.49% cannot be “fair and reasonable” under Duren; the black jury-service-eligible population of Allegheny County is large enough such that the troubling comparative disparity of 54.49% is probative of a Sixth Amendment violation. See id. at 242 (“[C]omparative disparity . . . is most useful when dealing with a group that comprises a large percentage of the population.”). The black jury-service-eligible population, however, is nonetheless small enough such that the absolute disparity of 5.83% in this case

15 “understates the systematic representative deficiencies.” Id. (quoting United States v. Shinault, 147 F.3d 1266, 1273 (10th Cir. 1998)). As discussed above, the absolute disparity in this case has an absolute maximum limit of 10.7%, which assumes complete exclusion of black persons from service on venires and a comparative disparity of 100%; thus, as illustrated by the Appendix, demanding a higher absolute disparity in this case would require a comparative disparity that would quickly approach 100% and complete exclusion. Therefore, underrepresentation of black persons on juries at a rate of 54.49% under these particular circumstances is sufficient to establish that such underrepresentation violates the Sixth Amendment’s fair-cross-section requirement.11 Cf. Garcia-

11 The unconstitutional nature of the underrepresentation of black persons on venires in Allegheny County comes into stark relief when one considers it in the broader context of the ultimate goal of the Supreme Court’s jurisprudence regarding racial discrimination in jury selection. As a result of Taylor v. Louisiana, 419 U.S. at 538, and its progeny (including Duren), the Supreme Court prohibits the state from discriminating on the basis of, among other things, race when compiling jury pools and assembling venires from which petit juries are drawn. See id. (“[J]ury wheels, pools of names, or venires from which juries are drawn must not systematically exclude distinctive groups in the community and thereby fail to be reasonably representative thereof.”). As a result of Strauder v. West Virginia, 100 U.S. 303, 305 (1880), and its progeny (including Batson v. Kentucky, 476 U.S. 79 (1986)), the Supreme Court prohibits the state from discriminating on the basis of race when selecting petit juries from those venires. See Batson, 476 U.S. at 86 (“The Equal Protection Clause guarantees the defendant that the State will

16 not exclude members of his race from the jury venire on account of race.” (citing Strauder, 100 U.S. at 305)). Although “a defendant has no right to a ‘petit jury composed in whole or in part of persons of his own race,’” id. at 85 (emphasis added) (quoting Strauder, 100 U.S. at 305), the upshot of Taylor and Strauder and their progeny is that a defendant’s petit jury should be reasonably representative of the racial demographics of her community because the empanelment of the petit jury should be the result of a process free from racial discrimination: venires cannot be assembled in a racially discriminatory way, and the state cannot select petit juries in a racially discriminatory way, and thus the resulting petit juries should be reasonably representative of the racial demographics of the community.

If black persons were represented on venires in Allegheny County in the early 2000s in equal proportion to their representation in the jury-service-eligible population as a whole (10.7%), assuming that petit juries were empaneled properly in a race-neutral manner, we would expect every single criminal petit jury in Allegheny County to have had at least one black juror. Specifically, we would expect each criminal petit jury of twelve to have, on average, 1.3 black jurors (10.7% of 12). In reality, utilizing Howell’s statistics and assuming again that petit juries were empaneled properly in a race-neutral manner, we expect that approximately 42% of criminal petit juries in Allegheny County had zero black jurors—like the jury that convicted Howell. Specifically, we expect that each criminal petit jury of twelve had, on average, 0.58 black jurors (4.87% of 12). The Constitution simply cannot tolerate such a wide disparity that results solely from the unrepresentativeness of venires.

17 Dorantes v. Warren, 801 F.3d 584, 600 (6th Cir. 2015) (“[T]he absolute disparity for African-Americans of 3.45% and corresponding 42% comparative disparity are sufficient to satisfy the Duren second prong.”).

B.

Finally, Howell has satisfied the third prong of the test in Duren: he has produced sufficient evidence to demonstrate that the underrepresentation of black persons on venires “is due to the systematic exclusion of this group in the jury-selection process.” 439 U.S. at 364 (emphasis added).

Under Duren, Howell need only demonstrate that the underrepresentation of black persons is “‘systematic’—that is, inherent in the particular jury-selection process utilized.” Id. at 366. In other words, Howell simply must prove that the underrepresentation of black persons was “due to the system by which juries were selected.” Id. at 367. The term “systematic exclusion,” however, does not connote “intentional discrimination”: “intentional discrimination need not . . . be shown to prove a Sixth Amendment fair cross section claim.” Weaver, 267 F.3d at 244 (citing Duren, 439 U.S. at 368 n.26 (contrasting equal-protection challenges, which require evidence of discriminatory intent, with Sixth Amendment fair- cross-section challenges, which require proof of only “systematic disproportion itself”)). “Under Duren, ‘systematic exclusion’ can be shown by a large discrepancy repeated over time such that the system must be said to bring about the underrepresentation.” Id. For example, the Supreme Court held in Duren that the petitioner’s statistical evidence, which “demonstrate[ed] that a large discrepancy occurred not just occasionally, but in every weekly venire” during an eight-

18 month study period, “manifestly indicate[d] that the cause of the underrepresentation was systematic.” 439 U.S. at 367.

The majority holds that Howell cannot demonstrate that the underrepresentation of black persons was “systematic” for three reasons: (1) the process by which venires were assembled was “facially neutral,” insofar as veniremembers were drawn from voter-registration lists and motor-vehicle records; (2) the six-month study of venires upon which Howell relies is not of a sufficient duration to support a finding of “systematic exclusion”; and (3) Allegheny County was engaged in “on-going efforts to improve the representativeness of jury lists,” which, according the majority, makes “it less likely that the data reflects that underrepresentation is due to a systematic exclusion in the jury process.”

I disagree with the premises of each of these points. First, by giving weight to the fact that venires are assembled from “facially neutral” sources, it appears that the majority is requiring Howell to produce evidence of racially discriminatory intent, which he is not required to produce under Duren to state a Sixth Amendment claim. See id. at 368 n.26; accord Weaver, 267 F.3d 244. According to the concurring opinion, because Allegheny County assembled its venires from two facially neutral sources—voter-registration lists and motor-vehicle records—Allegheny County’s “system [went] above and beyond what is constitutionally required.” What the concurring opinion fails to grasp is that the use of race-neutral sources in assembling venires is only what the Fourteenth Amendment requires: the Fourteenth Amendment forbids the government from intentionally discriminating on the basis of race in assembling venires or petit juries. See Strauder, 100 U.S. at 305. The Sixth Amendment, by contrast,

19 requires that “representation of [a distinctive] group in venires from which juries are selected [must be] fair and reasonable in relation to the number of such persons in the community.” Duren, 439 U.S. at 364 (quoting Taylor v. Louisiana, 419 U.S. 522, 538 (1975)). “[I]ntentional discrimination need not be shown to prove a Sixth Amendment fair[-]cross[-]section claim,” and thus the fact that Allegheny County assembled its venires from race-neutral sources is immaterial to Howell’s Sixth Amendment claim. Weaver, 267 F.3d at 244. The majority and concurring opinions thus disregard our observation in Weaver that “if the use of voter registration lists”—a facially neutral source—“over time did have the effect of sizeably underrepresenting a particular class or group of the jury venire, then under some circumstances, ‘this could constitute a violation of a defendant’s fair-cross-section rights under the [S]ixth [A]mendment.’” Id. at 244–45 (alteration in original) (quoting Bryant v. Wainwright, 686 F.2d 1373, 1378 n.4 (11th Cir. 1982)). This is not, as the concurring opinion phrases it, a “theoretical possibility”: Howell’s very statistics establish that the use of voter-registration lists and motor- vehicle records resulted in the underrepresentation of black persons on venires in Allegheny County at a rate of 54.49%, even though Allegheny County used race-neutral sources to assemble its venires.

Second, taken together with other evidence, the six- month duration of the study upon which Howell relies is sufficient to demonstrate that the underrepresentation of black persons was “systematic.” The six-month duration of the study in this case is sufficiently similar to the eight-month duration of the study in Duren, which, standing alone, “manifestly indicate[d] that the cause of the underrepresentation was systematic.” 439 U.S. at 367. Admittedly, Duren presented a

20 stronger set of facts, from which the Supreme Court could even “establish[] when in the selection process the systematic exclusion took place,” but nowhere in Duren does the Supreme Court hold that a litigant needs such a strong set of facts to prevail on a fair-cross-section claim; rather, the core holding of Duren in this regard is that a litigant must prove merely that the “cause of the underrepresentation was systematic—that is, inherent in the particular jury-selection process utilized”—and that a study with an eight-month duration “manifestly indicates” such a “systematic” cause. Id. Further, by relying on Ramseur for the proposition that a study with a duration of two years was not sufficient to demonstrate systematic underrepresentation, the majority disregards the fact that Ramseur is in direct conflict with Supreme Court precedent in Duren on this point, and Ramseur should not be considered good law in this regard. Indeed, our Court previously has noted that we undertook a flawed analytical approach in Ramseur with respect to the second and third prongs of Duren. See Weaver, 267 F.3d at 241 (“In our brief discussion of Ramseur’s Sixth Amendment claim, we appear to have combined the second and the third prongs of Duren . . . .”).

Third, contrary to the majority’s assertion, the evidence in this case that Allegheny County took steps to increase racial diversity on venires tends to suggest that the underrepresentation of black persons was systematic, not the opposite. The Jury Coordinator of the Allegheny County Court Administrator’s Office testified that “one of the parts of [his] mission ha[d] been to address concerns about the numbers of discrete races and colors . . . of people that [we]re represent[ed o]n our jury panels.” App. 137. The Jury Coordinator testified that “the most important” of his efforts to “address those concerns” was to “completely revise the questionnaire” that is

21 mailed to prospective jurors as part of the process of selecting veniremembers. Id. This amounts to an admission by Allegheny County that it knew that certain racial groups were underrepresented on venires and that the cause of the underrepresentation was the system by which veniremembers were selected because Allegheny County attempted to address the problem—and, indeed, eventually ameliorated the problem—by altering the system. This is not, as the majority asserts, evidence that undermines Howell’s case; this is evidence in Howell’s favor.

Therefore, Howell has satisfied the third prong of the test espoused in Duren. The six-month study upon which he relies is sufficiently similar in duration to the eight-month study in Duren such that the duration of the study indicates that the system of selecting potential jurors caused the underrepresentation, and the evidence with respect to Allegheny County’s attempts to alter the system to increase racial diversity suggest that Allegheny County itself believed the problem of underrepresentation was systematic.

III.

While I find that Howell’s statistics are reliable and help establish a prima facie violation of his Sixth Amendment fair- cross-section rights, the focus on and discussion of statistics and statistical concepts in this case—statistical reliability, the difference between standard deviation and standard error, the import of absolute disparity versus comparative disparity— obscures what is a relatively straightforward question: Did the process of selecting potential jurors result in the underrepresentation of black persons on venires in Allegheny County to a degree that is constitutionally unacceptable? In my view, the answer to that question must be “yes”: Howell

22 has demonstrated that black persons were underrepresented on venires to a troubling degree and that the underrepresentation was caused by the system of selecting prospective jurors, in violation of the Sixth Amendment’s fair-cross-section requirement.

There is evidence in the record to suggest that the court administrators in Allegheny County eventually implemented policies that remedied the underrepresentation of black persons on venires. The underrepresentation of black persons on venires, however, had not been remedied at the time of Howell’s trial, and, because Howell established that black persons were underrepresented on venires at an alarming rate, his Sixth Amendment right to have his petit jury drawn from a fair cross-section of the community was violated.

For the reasons stated above, I respectfully dissent. Because Howell has established a prima facie fair-cross- section violation, I would remand to the District Court to determine whether the Commonwealth can “justify[] this infringement by showing [that] attainment of a fair cross[-] section [was] incompatible with a significant state interest.” Duren, 439 U.S. at 368.

23 Appendix

Illustrative Absolute and Comparative Disparity Figures for Black Persons Serving on Venires in Allegheny County in the Early 2000s

(with Increases/Decreases in Venire Representation of 0.2%) (with Howell’s Statistical Evidence Shaded in Grey)

Percentage Percentage Absolute Comparative of of Venires Disparity Disparity Population 10.7% 10.7% 0.0% 0.00% 10.7% 10.5% 0.2% 1.87% 10.7% 10.3% 0.4% 3.74% 10.7% 10.1% 0.6% 5.61% 10.7% 9.9% 0.8% 7.48% 10.7% 9.7% 1.0% 9.35% 10.7% 9.5% 1.2% 11.21% 10.7% 9.3% 1.4% 13.08% 10.7% 9.1% 1.6% 14.95% 10.7% 8.9% 1.8% 16.82% 10.7% 8.7% 2.0% 18.69% 10.7% 8.5% 2.2% 20.56% 10.7% 8.3% 2.4% 22.43% 10.7% 8.1% 2.6% 24.30% 10.7% 7.9% 2.8% 26.17% 10.7% 7.7% 3.0% 28.04%

24 Percentage Percentage Absolute Comparative of of Venires Disparity Disparity Population 10.7% 7.5% 3.2% 29.91% 10.7% 7.3% 3.4% 31.78% 10.7% 7.1% 3.6% 33.64% 10.7% 6.9% 3.8% 35.51% 10.7% 6.7% 4.0% 37.38% 10.7% 6.5% 4.2% 39.25% 10.7% 6.3% 4.4% 41.12% 10.7% 6.1% 4.6% 42.99% 10.7% 5.9% 4.8% 44.86% 10.7% 5.7% 5.0% 46.73% 10.7% 5.5% 5.2% 48.60% 10.7% 5.3% 5.4% 50.47% 10.7% 5.1% 5.6% 52.34% 10.7% 4.9% 5.8% 54.21% 10.7% 4.87% 5.83% 54.49% 10.7% 4.7% 6.0% 56.07% 10.7% 4.5% 6.2% 57.94% 10.7% 4.3% 6.4% 59.81% 10.7% 4.1% 6.6% 61.68% 10.7% 3.9% 6.8% 63.55% 10.7% 3.7% 7.0% 65.42% 10.7% 3.5% 7.2% 67.29% 10.7% 3.3% 7.4% 69.16% 10.7% 3.1% 7.6% 71.03%

25 Percentage Percentage Absolute Comparative of of Venires Disparity Disparity Population 10.7% 2.9% 7.8% 72.90% 10.7% 2.7% 8.0% 74.77% 10.7% 2.5% 8.2% 76.64% 10.7% 2.3% 8.4% 78.50% 10.7% 2.1% 8.6% 80.37% 10.7% 1.9% 8.8% 82.24% 10.7% 1.7% 9.0% 84.11% 10.7% 1.5% 9.2% 85.98% 10.7% 1.3% 9.4% 87.85% 10.7% 1.1% 9.6% 89.72% 10.7% 0.9% 9.8% 91.59% 10.7% 0.7% 10.0% 93.46% 10.7% 0.5% 10.2% 95.33% 10.7% 0.3% 10.4% 97.20% 10.7% 0.1% 10.6% 99.07% 10.7% 0.0% 10.7% 100.00%