Albert Woodfox v. Burl Cain, Warden

Petitioner-Appellee Albert Woodfox is once again before this Court in connection with his federal habeas petition. The district court had originally granted Woodfox federal habeas relief on the basis of ineffective assistance of counsel, but we held that the district court erred in light of the deferential review afforded to state courts under the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”), and therefore vacated the district court’s decision. 1 We then remanded the case to the district court to consider the only remaining claim, which related to allegations of discrimination in the selection of the grand jury foreperson. 2 On remand, the district court held that the state court was not entitled to AEDPA deference; that Woodfox had successfully made out a prima fade case of discrimination in the se *363 lection of the grand jury foreperson; and that the State of Louisiana, acting through Respondent-Appellant Warden Burl Cain, had failed to rebut the prima facie case. 3 The district court once again granted federal habeas relief. 4

The State now appeals that grant of habeas relief. Because we find that AED-PA deference should not be granted, that Woodfox successfully made his prima facie case at the district eourt.level, and that the State failed in its rebuttal, we AFFIRM.

I

A

This case has a long and complicated factual and procedural history. Because of our detailed recitation of this history in our earlier opinion, we explain here only those facts relevant to the claim at issue: discrimination in the selection of the grand jury foreperson.

We begin with an important observation. Woodfox’s claim is not just about the selection of the grand jury foreperson. Rather, it is also about the selection of the grand jury itself. The grand jury system used for Woodfox’s re-indictment was the same as the one challenged in Campbell v. Louisiana, 5 As the Supreme Court explained, the Louisiana system of grand jury foreperson selection, at the time, was unlike most other systems. Under most systems, “the title ‘foreperson’ is bestowed on one of the existing grand jurors without any change in the grand jury’s composition.” 6 But under the Louisiana system at issue, “the judge select[ed] • the foreperson from the grand jury venire before the remaining [eleven] members of the grand jury [were] chosen by lot.” 7 The foreperson had the same voting power as all the other grand jurors. Thus, in effect, the judge chose one grand juror. This case then is one that alleges discrimination in the selection of the grand jurors, an important constitutional challenge. “For well over a century, the Supreme Court has held that a criminal conviction of an African-American cannot stand under the Equal Protection Clause of the Fourteenth Amendment if it is based on an indictment of a grand jury from which African-Americans were excluded on the basis of race.” 8

B

In 1972, Albert Woodfox was an inmate at the Louisiana State Penitentiary serving a fifty-year sentence for armed robbery. On April 17, 1972, the body of Brent Miller, a prison guard at the penitentiary, was found in a pool of blood, having been stabbed 32 times. Woodfox, along with three other prisoners, was identified as one of the assailants. Woodfox was tried twice for the murder. Initially, he was indicted in 1972 and convicted in 1973. That conviction was overturned in state court post-conviction proceedings. As a result, he was re-indicted in 1993 by a grand jury in West Feliciana Parish. The late Judge Wilson Ramshur of the 20th Judicial District appointed the grand jury’s foreperson. 9 Woodfox was convicted of sec *364 ond-degree murder in 1998. Woodfox was sentenced to life imprisonment, without the benefit of parole, probation, or suspension of sentence in February 1999.

After his re-indictment, Woodfox moved to quash the new indictment based upon allegations of discrimination in the selection of the grand jury foreperson. The state trial court denied this motion. After his second conviction, on direct appeal, Woodfox raised several issues, including the trial court’s denial of the motion to quash the indictment. On June 23, 2000, the Louisiana Court of Appeal, First Circuit affirmed the conviction and sentence, 10 and in doing so, held that the trial court made no error in denying the motion to quash. The Louisiana First Circuit found that the claim about discrimination in the selection of the grand jury foreperson failed because Woodfox did not successfully establish a prima facie case. According to the Louisiana First Circuit, Woodfox had not shown “substantial underrepresentation of his race.” Woodfox is African-American. The evidence available to the Louisiana First Circuit demonstrated that between March 1980 and March 1995, African-Americans constituted 44% of all registered voters in the Parish, while constituting only 27% of all grand jury forepersons. First, the Louisiana First Circuit did not think this disparity was large enough. Second, the court held that the percentage of African-American registered voters did “not indicate how many were qualified to serve as grand jurors.” 11 The court reasoned that the difference could have been reduced, if not eliminated, if eligible population statistics instead of gross population statistics had been used. Woodfox filed a -writ application with the Louisiana Supreme Court, which.was denied on June 15, 2001, and then filed a writ of certiorari with the United States Supreme Court, which was denied on Novem-. ber 13, 2001. 12

C

After failing to gain relief on direct appeal, Woodfox next filed his application for state post-conviction relief. He raised several claims, including the claim regarding discrimination in the selection of the grand jury foreperson. In support of that claim, Woodfox produced new evidence. First, Woodfox presented the disparity over a longer period of time. Between 1970 and 1990, African-Americans represented between 40%-56% of the non-incarcerated population of the Parish. But, between 1964 and 1993, African-Americans represented only 12% of all grand jury forepersons. Second, in response to the earlier decision on direct appeal, Woodfox presented the disparity using eligible population statistics, instead of general population statistics. For the eligible population statistics, Woodfox chose to rely on the race percentages found within the grand jurors drawn by lot, i.e., the racial makeup of non-foreperson grand jurors. 13 Woodfox compiled the race data *365 with information he gathered with assistance from the registrar of voters in the Parish, and he presented the data to the extent he could determine the race of all the non-foreperson grand jurors. Between 1964 and 1993, African-Americans constituted an average of 36% of the non-foreperson grand jurors. During the same period, as mentioned above, African-Americans represented only 12% of all grand jury forepersons. 14

The State filed a response to this application for state post-conviction relief. 15 In its answer, the State urged the rejection of the grand jury foreperson discrimination claim. The State argued that the new evidence was essentially the same as the evidence presented on direct appeal, except that the time period had been changed from 1980-1995 to 1964-1993. The State also argued that the new evidence, which presented the race of the non-foreperson grand jurors was publicly available information that the defense could have presented during direct appeal but did not. As a result, the State argued that the claim was “meritless,” that the matter had already been ruled upon, and that the state post-conviction court need not revisit the issue.

On October 25, 2004, the 21st Judicial District Court sitting as the state post-conviction court denied the application for post-conviction relief. The state post-conviction court’s decision was comprised of two separate documents: a “Judgment” and a statement of ‘Written Reasons.”

In the “Judgment,” the state post-conviction court denied Woodfox’s application in entirety, stating that the application was “fully addressed” by the State’s answer and that “[a] review of the record of these proceedings, as well as the answer, indicates that there is no need to hold an evidentiary hearing in these proceedings. For written reasons this day adopted and assigned, the Court finds that the allegations are without merits and the Application may be denied without the necessity of further proceedings.”

In the “Written Reasons,” the state post-conviction court noted that Woodfox had to bear the burden of proving that he was entitled to habeas relief. It then cited to the Louisiana Code of Criminal Procedure article 930.2, and then stated: “In light of such burden of proof, the Court has fully considered the application, the answer, and all relevant documents and has determined that Petitioner has . failed to carry his burden of proof. In determining that Petitioner’s application should be denied, the Court, moreover, adopts the State’s [answer] as the written reasons for the Court’s decision.”

After failing to get relief from the state post-conviction court, Woodfox filed a writ application with the Louisiana First Circuit, which was denied on August 8, 2005. He then filed a writ application with the Louisiana Supreme Court, which was denied on September 29, 2006.

*366 D

Woodfox timely filed his petition for federal habeas relief pursuant to 28 U.S.C. § 2254 on October 11, 2006 and amended it on February 14, 2007. Woodfox made several claims for habeas relief, including claims of ineffective assistance of counsel, claims of suppression of exculpatory evidence, and the claim of discrimination in the selection of the grand jury foreperson.

The case was referred to a magistrate judge. As to the ineffective assistance of counsel claims, the magistrate judge found that Woodfox’s 1998 trial counsel had performed deficiently in some respects and thus prejudiced Woodfox, and therefore recommended that the conviction be vacated and the case remanded to state court. 16 As to the grand jury foreperson discrimination claim, the magistrate judge ruled in the alternative. The magistrate judge found that Woodfox had presented evidence sufficient to support a prima facie ease of discrimination, but that an evidentiary hearing would be necessary to allow the State an opportunity to rebut the prima facie case. But the magistrate judge did not conduct the hearing because Wood-fox’s ineffective assistance claims were sufficient to overturn his conviction. Instead, the magistrate judge recommended that if the district judge disagreed with the resolution of the ineffective assistance claims, then the matter be referred back for the evidentiary hearing.

On July 8, 2008, the district court adopted the magistrate judge’s report and granted the writ of habeas corpus. The State filed a motion to supplement the record and a motion to reconsider. On September 11, 2008, the district court reaffirmed its July 8th ruling granting the writ of habeas corpus. The State appealed the grant of habeas corpus. As discussed above, we vacated the district court’s judgment based upon the highly deferential review mandated by AEDPA. 17 But the claim of discrimination in the selection of the grand jury foreperson was not before us, 18 and we remanded for the resolution of this remaining claim. 19

E

Upon remand, the district court first held that the state court’s decision — specifically the Louisiana First Circuit’s June 23rd ruling — was an unreasonable application of clearly established law as determined by the Supreme Court and therefore should not be afforded AEDPA deference. ' It then held an evidentiary hearing on May 29-31, 2012. 20

The district court ruled that the relevant time period for grand jury foreperson selection in West Feliciana Parish was 1980 through March 1993. 21 To establish his prima facie claim, Woodfox used both general and eligible population statistics. First, the general population statistics showed that in 1990, the percentage of African-Americans in the Parish, excluding prisoners, was 44%. 22 The percentage of African-Americans among registered *367 voters between 1980 and 1993 was 43.5%. 23 Second, the eligible population statistics showed that between 1980 and March 1993, there were 297 non-foreperson grand jurors; Woodfox was able to establish the race of 277 of these grand jurors. 24 Only 113 out of 277 non-foreperson grand jurors were African-American, or 40.8%. 25 Third, during this time, only 5 out of 27 grand jury forepersons were African-American, or 18.5%. 26 Based on this and other factors, the district court found that Woodfox had successfully made out a prima facie case. 27 The district court then rejected the State’s rebuttal case, which included statistical evidence that aimed to discredit the prima facie case as well as evidence attempting to demonstrate that West Feliciana Parish judges relied on racially neutral criteria in selecting the grand jury foreperson. 28 The district court granted habeas relief. 29 The State now appeals.

II

“In a habeas corpus appeal, we review the district court’s findings of fact for clear error and its conclusions of law de novo, applying the same standards to the state court’s decision as did the district court.” 30 Under 28 U.S.C. § 2254(d), we cannot grant a writ of habeas corpus with respect to any claim adjudicated on the merits in state court unless such adjudication:

(1) resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States; or

(2) resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding. 31

For a challenge to a state court decision under § 2254(d)(1), the Supreme Court has clarified that the “contrary to” inquiry is different from the “unreasonable application” inquiry. 32 A state court’s decision is “contrary to” clearly established federal law if “the state court arrives at a conclusion opposite to that reached by [the Supreme Court] on a question of law or if the state court decides a case differently than [the Supreme Court]'has on a set of materially indistinguishable facts.” 33 A state court’s decision involves an “unreasonable application” of clearly established federal law if “the state court identifies the correct governing legal principle from [the Supreme Court’s] decisions but unreasonably applies that principle to the facts of the prisoner’s case.” 34 In reviewing a state court’s decision under the “unreasonable application” prong, we focus on “the ultimate legal conclusion that the state court reached and not on whether the state court considered and discussed every an *368 gle of the evidence.” 35 The Supreme Court has clarified that when a claim is adjudicated on the merits, for the purposes of review under § 2254(d)(1), the record is limited to the one before the state court, even'if the state court issued a summary affirmance. 36

A challenge to a state court decision under § 2254(d)(2) challenges the determination of. facts by the state court. 37 Under 28 U.S.C. § 2254(e)(1), “a determination of a factual issue made by a State court shall be presumed to be correct” and the habeas petitioner “shall have the burden of rebutting the presumption of correctness by clear and convincing evidence.” 38 Section 2254(e)(1) is the “arguably more deferential standard.” 39 The Supreme Court has recognized a division among the circuits on the interplay betwéen these two statutory provisions, 40 but has yet to resolve this question. 41 Regardless, a .state court’s factual determination is “not unreasonable merely because the federal habeas court would have reached a different conclusion in the first instance.” 42 For claims that are not adjudicated on the merits in the state court, we apply a de novo standard of review. 43

Finally, “whether the grand jury was selected in a systematically unrepresentative or racially discriminatory manner, has long been recognized to be a question of law or a mixed question of fact and law.” 44

III

The first issue in this appeal is which state court decision ought to be examined for AEDPA deference. The State argues that it is the Louisiana First Circuit’s June 23rd ruling on direct appeal which should be examined. Indeed, the district court examined this ruling for AEDPA deference. Woodfox argues that the state post-conviction court’s October 25th ruling should be examined. 45

*369 Under AEDPA, “we review the last reasoned state court decision.” 46 Using the “look through” doctrine, we “ignore — and hence, look through — an unexplained state court denial and evaluate the last reasoned state court decision.” 47 In Ylst v. Nunnemaker 48 on direct appeal, the state appeals court had applied a procedural bar to a claim. 49 The petitioner subsequently filed a petition for habeas corpus with the state supreme court, “invoking the original jurisdiction” of that court. 50 That petition was denied without opinion. 51 In holding that the procedural bar was still valid, the Supreme Court applied a presumption that “[w]here there has been one reasoned state judgment rejecting a federal claim, later unexplained orders upholding that judgment or rejecting the same claim rest upon the same ground.” 52 Ylst also made clear that if the later state court decides the question differently than the original state court, then the later judgment has controlling effect. 53

Here, working backwards through the state adjudicatory process, it is clear that during state post-conviction proceedings, neither the Louisiana First Circuit nor the Louisiana Supreme Court issued a reasoned opinion. At the very least, then, we have to examine the state post-conviction court’s October 25th ruling. But the State contends that as to the grand jury foreperson discrimination claim, the October 25th ruling by the state post-conviction court was not on an adjudication on the merits. The State contends that the state post-conviction court applied a special type of bar: Louisiana Code of Criminal Procedure article 930.4(A), which states that “[u]nless required in the interest of justice, any claim for relief which was fully litigated in an appeal from the proceedings leading to the judgment of conviction and sentence shall not be considered.” 54 As we have recognized before, “[t]he bar imposed by article 930.4(A) is not a procedural bar in the traditional sense, nor is it a decision on the merits.” 55 The State argues that the state post-conviction decision cannot be examined for AEDPA deference because it neither adjudicated the claim on the merits nor applied a procedural bar in the traditional sense. The State wishes us to look even further back to the opinions on direct appeal. Specifically, the State argues that the Louisiana First Circuit’s June 23rd decision on direct appeal is the only one that adjudicated this claim on the merits; that opinion should be examined for AED-PA deference. The upshot of this argument is clear. The Louisiana First Circuit rejected Woodfox’s claim because he had failed to present eligible population statistics. Thus, the § 2254(d) inquiry would ask whether the state court’s opinion was *370 contrary to or an unreasonable application of clearly established federal law as determined' by the Supreme Court in requiring eligible population statistics. By contrast, Woodfox did present eligible population statistics to the state post-conviction court. Thus, the § 2254(d) inquiry would ask whether the state court’s opinion was contrary to or an unreasonable application of deafly established federal law in rejecting the disparity demonstrated.

To our eyes, the state-post conviction opinion was an adjudication on the merits and should be examined for AEDPA deference. This conclusion is the product of two different reasons. First, the law-of-the-case doctrine suggests that this was a merits adjudication. “The law-of-the-case doctrine posits that when a. court decides upon a rule of law, that decision should continue to govern the same issue in subsequent stages in the same case.” 56 “[A]n issue of fact or law decided on appeal may not be reexamined either by the district court on remand or by the appellate court on subsequent appeal.” 57 During his first appeal to our Court, we specifically noted that the grand jury foreperson discrimination claim was not at issue. Yet when deciding the nature of the state-post conviction opinion we also held that “it is clear that the state [post-conviction] court decided all of Wood-fox’s claims on the merits.” 58 This holding binds us, and compels the conclusion that the state post-conviction court adjudicated the present claim on the merits.

Second, even if we reject the use of the law-of-the-case doctrine, we would still hold that the state post-conviction court adjudicated the claim on the merits. The Supreme Court clarified in Harrington v. Richter, 59 that “[w]hen a federal claim has been presented to a state court and the state court has denied relief, it may be presumed that the state court adjudicated the claim on the merits in the absence of any indication or state-law procedural principles to the contrary.” 60 The Richter presumption applies even where the habeas petitioner raises a federal claim and the “state court rules against the defendant and issues an opinion that addresses some issues but does not expressly address the federal claim in question.” 61 But the “presumption may be overcome when there is reason to think some other explanation for the state court’s decision is more likely.” 62 The presumption could be rebutted “either by the habeas petitioner (for the purpose of showing that the claim should be considered by the federal court de novo) or by the State (for the purpose of showing that the federal claim should be regarded as procedurally defaulted).” 63 For example, “a federal claim [that] is rejected as a result of sheer inadvertence,” would not be afforded the Richter presumption. 64 . Thus, we must presume that the state post-conviction opinion was an adjudication on the merits as to the grand *371 jury foreperson discrimination claim. And it is the State’s burden to demonstrate that a bar — such as Article 930.4(A) — -was applied. The State simply cannot carry this burden.

We have adopted a .three-part test when it is unclear whether a state court’s opinions adjudicates a claim on the merits. We consider:

(1) what the state courts have done in similar cases;

(2) whether the history of the ease suggests that the state court was aware of any ground for not adjudicating the case on the merits; and

(3) whether the state courts’ opinions suggest reliance upon procedural grounds rather than a determination on the merits. 65

As to the first prong, as we noted in Woodfox’s first appeal, the state post-conviction court held that Woodfox’s claims had no merit and that it would adopt the State’s answer. The court cited Louisiana Code of Criminal Procedure article 930.2, which provides that “[t]he petitioner in an application for post conviction relief shall have the burden of proving that relief should be granted.” 66 The Louisiana Supreme Court cites Article 930.2 both in cases where the petitioner has failed to carry his burden on the merits and where the petitioner has failed to meet his burden on some procedural point. 67 Moreover, the Louisiana Courts of Appeals have repeatedly cited Article 930.4(A) when relying upon it, while in this case no such citation was made. 68 Thus, consideration of what the state courts have done in similar cases does not support overcoming the presumption that the state court here issued a decision on the merits.

As to the second prong, the history of the case suggests that the state court was aware of a possible ground for not adjudicating the case on the merits. The State primarily relies on the answer that it submitted to the state post-conviction court. The State argued that the new evidence presented was both untimely and substantially similar to evidence already considered on appeal, and thus did not justify revisiting the already-litigated issue. This reasoning could support a merits decision: it urges that the logic behind the merits decision on appeal retained its force because nothing of consequence had been added in the post-conviction case. Indeed, the answer explicitly asserted that the claim was “meritless.” On the other hand, though it never cited Article 930.4(A), the State’s argument could also provide grounds supporting a non-merits decision based on that Article. It is worth noting that a distinction may be drawn between the state court being “aware of any ground for” a non-merits decision and the court being aware simply of the argument that such a ground exists. Putting aside that distinction, however, it does appear that the court was aware of a ground that might have supported a non-merits decision under Article 930.4(A).

*372 As to the third prong, we find ourselves constrained to follow the logic adopted in Woodfox’s earlier appeal. We inquire whether the state post-conviction court’s opinion suggests reliance upon procedural grounds rather than a determination of the merits. In its “Judgment,” the court stated that the record along with the State’s answer indicated that, “the allegations are without merit.” In its “Written Reasons,” the court stated that it had considered “the application, the answer, and all relevant documents” before concluding that Woodfox failed to meet his burden. The court then stated that “moreover” it was adopting the State’s answer. As we noted in Woodfox’s earlier appeal and note again now, “moreover” means “[i]n addition thereto, also, furthermore, likewise, beyond this, beside this,” 69 or “in addition to what has been said.” 70 Resultantly, the state post-conviction court reviewed the. record in its entirety and found no merit as to any of Woodfox’s claims. In addition to this conclusion, the court also adopted the State’s answer which, as discussed above, could support either a merits or non-merits decision.

We cannot simply assume that there was an implicit application of the Article 930.4(A) bar. To do so would fly directly in the face of the presumption of merits adjudication the Supreme Court has clearly announced. In this case, the factors on balance point to the conclusion that the state post-conviction court adjudicated the grand jury foreperson discrimination claim on the merits. Therefore, the district court erred in examining afresh the Louisiana First Circuit ruling. We now turn to examine the state post-conviction decision, according the deference required by AED-PA.

IV

If the state post-conviction opinion withstands the scrutiny of §. 2254(d), thereby affording AEDPA deference, habeas relief may not be granted.

A

In Castaneda v. Partida, 71 the Supreme Court held that to show that an equal protection violation has occurred in a grand jury context, the “defendant must show that the procedure- employed resulted in substantial underrepresentation of his race or of the identifiable group to which he belongs.” 72 To make a prima facie case, the petitioner must do three things:

The first step is to establish that the group is one that is a recognizable, distinct class, singled out for different treatment under the laws, as written or as applied. Next, the degree of under-representation must be proved, by comparing the proportion of the group in the total population to the proportion called to serve as grand jurors, over a significant period of time. This method of proof, sometimes called the ‘rule of exclusion,’ has been held to be available as a method of proving discrimination in jury selection against a delineated class. Finally, as noted above, a selection procedure that is susceptible of abuse or is' not racially neutral supports the presumption of discrimination raised by the statistical showing. 73

*373 Upon showing of this prima facie case, “the burden then shifts to the State to rebut that case.” 74

There can be no dispute that the first and third elements of the prima facie case have been met. African-Americans are a distinct, cognizable class that have been singled out for discrimination. 75 Next, both federal and state courts have recognized that the system of selecting the grand jury foreperson then in place was susceptible to abuse.- 76 Indeed, as the Louisiana Supreme Court held before Woodfox’s state post-conviction proceedings, the system “was unquestionably subject to abuse according to subjective criteria that may include race and gender.” 77 If the state post-conviction court had rejected- the prima facie case on either of these prongs, its determination would have clearly been contrary to or an unreasonable application of clearly established federal law. Therefore, the state post-conviction court could only have rejected this claim based on the second element: that the degree of underrepresentation had not been proven over a significant period of time.

B

In making our § 2254(d) inquiry, we begin first by clarifying a question we are not answering. We need not decide the question of whether a state court errs when it requires eligible population statistics rather than general population statistics from a petitioner in making out a prima facie case. That issue is quite complicated. To begin, Castaneda allowed the use of general population statistics in proving the degree of underrepresentation. Even though Chief Justice Burger argued in dissent that “eligible population statistics, not gross population figures, provide the relevant starting point,” 78 the majority rejected this position. The majority found that the petitioner had made a prima facie case, thus shifting the burden of rebuttal to the State. 79 Next, the Castaneda Court faulted the Texas state court under review for speculating on its own motion that general population statistics were not reliable, and requiring the use of eligible population statistics. 80 Instead, Castaneda made it the State’s burden to show that the statistical disparities are unreliable through the us.e of eligible population statistics. 81 Thus, Castaneda stands for the proposition that petitioners can always prevail on the prima facie case using general population statistics, and it is the State’s burden to produce eligible population statistics.

But Castaneda’s holding is also limited by its context. First, Castaneda compares the general population statistics to a group of persons not at issue in this case: people called to serve as grand jurors, not those who actually served as grand jurors. As the Supreme Court explained at the time, the Texas method of selecting grand jurors was unique. A Texas state district judge would appoint jury commissioners; those *374 jury commissioners would in turn select a list of 15 to 20 people from which the grand jury would eventually be drawn. 82 When at least 12 of those people appeared appear in court, the district judge would proceed to test their qualifications. 83 Thus, “qualifications [were] not tested until the persons on the list appeared] in the court.” 84 Castaneda compares the general population statistics to those called by the jury commissioners. In other words, it compares population statistics to a group that had not yet been qualified. By contrast, Woodfox attempted to compare his population statistics to persons who actually served as grand jury forepersons, i.e., a group of qualified persons. Second, the Supreme Court also explained that it preferred not to use eligible population statistics because the idea that eligible population statistics ought to be used was not brought up until oral argument: “[T]here are so many implicit assumptions in this analysis, and we consider it inappropriate for us, as an appellate tribunal, to undertake this kind of inquiry without a record below in which those assumptions were tested.” 85

Further complicating the question is our decision in United States ex rel. Barksdale v. Blackburn. 86 In that case, the “issue [was] whether general population statistics or more meaningful eligible population statistics should be used where ... those statistics are in the record.” 87 We acknowledged that Castaneda used general population statistics, but held that Castaneda “should not be read to require using those figures.” 88 We decided that “statistics describing the presumptively eligible black juror population, rather than the general black population, provide the proper starting point for an inquiry into racial disparities in the Parish.” 89 This was because such “appropriate statistics had been developed in the'record.” 90

Since Woodfox presented both general and eligible population statistics to the state post-conviction court, however, our § 2254(d) inquiry is much simpler. We simply have to ask whether the state post-conviction court’s rejection of the statistics presented was contrary to or an unreasonable application of clearly established federal law as determined by the Supreme Court.

C

Recall that Woodfox presented the following information to the state post-conviction court. First, that between 1970 and 1990, African-Americans represented between 40%-56% of the non-incarcerated population of the Parish. Second, that between 1964 and 1993, African-Americans constituted an average of 36% of the non-foreperson grand jurors. 91 This constituted his proof of general and eligible population statistics. Third, that between 1964 and 1993, African-Americans represented only 12% of all grand jury fore *375 persons. Therefore, using the low end of general population statistics, the absolute disparity would have been 28%, and using the eligible population statistics it would have been 24%.

State courts are not restricted to using only absolute disparity evidence to evaluate a prima facie case. 92 - However, absolute disparity evidence was the only kind of evidence put before the state post-conviction court in this case. The Supreme Court has provided useful indicators as to the amount of absolute disparity that is sufficient to satisfy the second element of the prima facie case. To begin, the Court has held- that underrepresentation by as much as 10% does not show purposeful discrimination based on race. 93 Next, in Castaneda, the petitioner successfully made his prima facie case by showing that Mexican-Americans constituted 79.1% of the county, yet constituted only 39% of those summoned for grand jury service: an absolute disparity of 40%. 94 Not only that, but Castaneda also highlighted the other absolute disparities that were acceptable to establish a prima facie case. For example, the Supreme Court has specifically allowed the following disparities to make out a prima facie case of grand jury discrimination: 14.7% 95 ; 18% 96 ; 19.7% 97 ; 23%. 98

Based on these figures, it is apparent that the absolute disparities before the state post-conviction court — either 24% or 28% — could not have been rejected without being an unreasonable application of federal law as determined by-the Supreme Court. These disparities are well within the range considered significant by the Supreme Court. As a result, the state post-conviction opinion cannot be afforded AEDPA deference under the § 2254(d) standard.

V

Having held that AEDPA deference is not warranted, we now turn to the proceedings held before the district court at the federal evidentiary hearing. We begin with the prima facie case made before the district court.

As a reminder, under Castaneda, there are three elements to the prima facie case: 1) the group has to be a recognizable, distinct class, singled out for different treatment under the laws, as written or as *376 applied, 2) the degree of underrepresentation must be proved' over a significant period of time, and 3) the selection procedure must be susceptible of abuse or must not be racially neutral. 99 Again, there can be no doubt that Woodfox met the first and third elements. Woodfox is African-American and African-Americans constitute a distinct, cognizable class. 100 Next, the Louisiana procedure for selecting grand jury forepersons prior to 1999 was “unquestionably subject to abuse according to subjective criteria that may include race and gender.” 101

As to the second element, the district court held that the relevant time period was between 1980 and March 1993. Recall that, to establish his prima facie claim, Woodfox used both general and eligible population statistics. First, the general population statistics showed that in 1990, the percentage of African-Americans in the Parish, excluding prisoners, was 44%. 102 The percentage of African-Americans among registered voters between 1980 and 1993 was 43.5%. 103 Second, the eligible population statistics showed that between 1980 and March 1993, there were 297 non-foreperson grand jurors; Woodfox was able to establish the race of 277 of these grand jurors. 104 Only 113 out of 277 non-foreperson grand jurors were African-American, or 40.8%. 105 Third, during this time, only 5 out of 27 grand jury forepersons were African-American, or 18.5%. 106 Based on these statistics, the district court found that a prima facie case had been established. We agree. The absolute disparity using general population statistics is at least 25%. This in itself would be enough to establish the prima facie case. First, our Court has previously allowed the use of general population statistics for this purpose. 107 Second, this disparity is exactly in the range the Supreme Court has found sufficient for a prima facie case. 108 Fifth Circuit precedent confirms that these numbers are enough. 109 Moreover, the absolute disparity using eligible statistics is 22.3%. 110 The district court did not err in finding that Woodfox had made out his prima facie case.

VI

The prima facie case made by Woodfox “therefore shifted the burden of proof to the State to dispel the inference of inten *377 tional discrimination.” 111 Before proceeding to the rebuttal case, however, we deal. with the evidentiary stages the district court set up for the proceedings.

The district court split its hearing into three stages. Stage One was to be the prima facie case. The prima facie case, as discussed above, was for between 1980 and March 1993. And it covered grand jury foreperson selections for all of West Feliciana Parish, i.e., it covered grand jury foreperson selections by both Judge Ramshur (the appointing judge in Woodfox’s re-indictment) and Judge William Kline. Stage Two was to be the State’s rebuttal, both as to statistics and race-neutral criteria used in the selection of grand jury forepersons. Stage Three was to' be' Woodfox’s reply because once the rebuttal was successful, the presumption of discrimination would disappear and Woodfox would again have the burden of showing discriminatory intent on the part of Judge Ramshur. This framework parallels the Supreme Court’s Batson 112 framework, which “(1) requires] defendants to establish a prima facie case of discrimination, (2) ask[s] prosecutors then to offer a race-neutral explanation ■ for their use of the peremptory,' and then (3) require[es] defendants to prove that the neutral reason offered is pretextual.” 113

The State argues that this three-stage process is not allowed under Castaneda; that although the district court pronounced that it did not need to reach Stage Three, it implicitly did so because it used Wood-fox’s eligible population statistics as the proper baseline, which were developed in Stage Three. While we agree, we do not find any reversible error. First, the Bat-son framework is not an exact analogy to the Castaneda framework. While in Bat-son a simple articulation of any race-neutral reason moves the process to the next stage, we have found the rebuttal stage of Castaneda to encompass more: it is an examination into whether there was intentional discrimination. 114 Thus, in Castaneda challenges, Stages Two and Three are really one and the same. Second, it is evident that the district court had to reach the evidence in Stage Three. As we discuss below, the State provided a statistical rebuttal and Woodfox a statistical .reply. For the district court to rely on Woodfox’s statistical reply, it necessarily reached what it termed Stage Three. This presents no reversible error. Even considering all the Stage Two and Three evidence together, the State fails in its rebuttal case.

VII

In rebuttal of the prima facie case, the State renews its arguments that the eligible population statistics used by Woodfox were not appropriate and that, in any case, the statistical disparity was not enough.

A

As Castaneda suggests, the State in rebuttal tried to introduce its own eligible population statistics. The State introduced an expert according to whom the eligible population statistics for the Parish showed that African-Americans were only 36.62% of the population eligible for grand jury foreperson service. The district court rejected the use of this figure, concluding *378 that the appropriate baseline for comparison was 40.8% from Woodfox’s eligible population statistics. We agree.

To understand the problematic nature of the State’s 36.62% baseline, it is important to understand how it was derived. The State’s expert started with the voter rolls for West Feliciana Parish. He then proceeded to screen out those people on the voter rolls who would be ineligible to serve as grand jurors, and did so by using illiteracy as his screening factor. But public records only contained the illiteracy data for 1980-1985 and 1988-1993. Moreover, only the data from 1980-1985 were broken down by race, and they indicated that 97.8% to 98% of illiterate voters were African-American. The expert used the smaller of these numbers (97.8%) and applied it to the 1988-1993 data to derive the percentage of illiterate voters who were African-American. Then, for the missing time period of 1986-1987, he used a regression analysis to determine the number of illiterate voters in the two year period. Finally, he then combined this illiteracy data with the voter rolls to conclude that African-Americans were only' 36.62% of the eligible population. In sum, this analysis relies on limited information about literacy rates from 1980-1985 and no evidence of literacy rates, broken down by race, from 1985-1993. Given this incomplete picture, we cannot find reversible error in the district court’s refusal to rely upon it. In Barksdale, for example, we similarly rejected the opinion of a state’s expert footed on his statistical analysis where he was “overzealous in his adjustment of the eligible population.” 115

The type of eligible population statistics provided by Woodfox have already been accepted by the Louisiana Supreme Court. First, the Louisiana Supreme Court has held that for the purposes of a grand jury foreperson discrimination claim, a petitioner can use the percentage of a racial group from the non-foreperson grand jurors as representative of eligible population statistics. 116 As to the State’s argument that Woodfox’s eligible population statistics are merely a sample and not the whole population, “common sense tells us that the group of grand jurors who actually served is ... a randomly-selected sample or subset of’ the eligible population. 117 More importantly, the district court concluded that the State’s statistics “relied on more incomplete data” than the statistics relied on by Woodfox. 118 It further found that “the State has altered the numbers to reduce the baseline of eligible African-Americans.” 119 Given the fact-intensive nature of the competing statistical inquiries and the district court’s through review of these questions, we hold the district court did not clearly err in finding that the appropriate baseline for eligible population statistics was 40.8%.

B

The State also argues that Woodfox failed to show statistical significance in the disparity. As both parties acknowledge, there are other statistical methods besides absolute disparity, such as disparity in standard deviations as well as hypothesis testing (including one-tailed and two-tailed testing).

We begin with a necessary observation. While it is true that Castaneda discussed disparity in standard deviations, the Su *379 preme Court conducted its analysis in absolute disparity terms by holding that a 40% disparity was enough to establish a prima facie case. 120 Indeed, we have “referred to statistical methods other than absolute disparity, but have never found a constitutional violation based on the data produced by such methods.” 121 In this case, the absolute disparities shown are within the range traditionally accepted by the Supreme Court tó establish a prima facie case of discrimination.

However, the gravamen of the State’s argument is that the absolute disparities shown are meaningless because they are statistically insignificant. The State argues that under the 40.8% baseline of eligible population statistics, the disparity is only 2.37 standard deviations. 122 Woodfox argues that standard deviations are not the appropriate method of measuring statistical significance. He argues next that under the more accurate one-tailed and two-tailed testing, he has shown statistical significance for both baselines.

We begin first with the disparity in standard deviations. The source of the difficulty is the Supreme Court’s general language in Castaneda, offering a description of standard deviation, but not an explanation of its context or use with regard to binomial distributions. As the Court explained:

If the jurors were drawn randomly from the general population, then the number of Mexican-Americans in the sample could be modeled by a binomial distribution. Given that 79.1% of the population is Mexican-American, the expected number of Mexican-Americans among the 870 persons summoned to serve as grand jurors over the 11-year period is approximately 688. The observed number is 339. Of course, in any given drawing some fluctuation from the expected number is predicted. The important point, however, is that the statistical model shows that the results of a random drawing are likely to fall in the vicinity of the expected value. The measure of the predicted fluctuations from the expected value is the standard deviation, defined for the binomial distribution as the square root of the product of the total number in the sample (here 870) times the probability of selecting a Mexican-American (0.791) times the probability of selecting a non-MexieanAmerican (0.209). Thus, in this case the standard deviation is approximately 12. As a general rule for such large samples, if the difference between ^the expected value and the observed number is greater than two or three standard deviations, then the hypothesis that the jury drawing was random would be suspect to a social scientist. The 11-year data here reflect a difference between the expected and observed number of Mexican-Americans of approximately 29 standard deviations. 123

Despite its generality, two important lessons are fairly drawn from this discussion. *380 First, in Castaneda, the difference between the expected and observed number was 29 standard deviations, very different from the 2.37 standard deviations present in this case. Second, and importantly, the Supreme Court did not define the number of standard deviations necessary to offer a statistically significant result. Instead, it observed only that a difference greater than 2 or 3 standard deviations would cause a social scientist to doubt that the difference had occurred by chance. This is important because the State primarily argues that a standard deviation between 2 and 3 is a “gray zone,” not necessarily implicating statistically significance. Since the disparity is only 2.37 standard deviations, the State argues that Woodfox has not shown statistical significance.

We need not linger further here because the district court found the one-tailed and two-tailed tests more appropriate and addressed the statistical significance issue in those terms. Woodfox’s expert explained that standard deviation is a crude tool to analyze symmetric, bell-shaped, normally-distributed data, but does not work where, as here, the data is not symmetrically distributed. Again, given the fact-intensive nature of the statistical inquiry, we can find no clear error in the district court’s opting to use the one-tailed and two-tailed tests.

The basics of hypothesis testing (including one-tailed and two-tailed tests) are explained through two simple examples. First, suppose that 50% of the population eligible to serve as jurors in a county ‘are women. 124 A jury is drawn from a panel of 350 persons selected by the clerk of the court, but the panel includes only 102 women, i.e., less than 50%. 125 Hypothesis testing answers the question of whether the shortfall in women can be explained by the mere play of random chance. 126 A statistician would formulate and test a null hypothesis, which in this case would see the panel of 350 as 350 persons drawn at random from the larger eligible population. 127 The expected number of women would be 50% of 350, which is 175. 128 The observed number is obviously less: 102. 129 The shortfall is the difference between 175 and 102: 73. 130 Hypothesis testing answers the question of how likely it is to find this disparity between the numbers— the probability is called the p-value. 131 “Large p-values indicate that a disparity can easily be explained by the play of chance.” 132 “[I]f p is very small, something other than chance must be involved.” 133 “In practice, statistical analysts typically use levels of 5% and 1%” for statistical significance. 134

Second, to demonstrate the difference between one-tailed and two-tailed testing, suppose a coin is tossed 1000 times and the result is 532 heads. 135 “The null hypothesis to- be tested asserts that the. coin is *381 fair.” 136 If correct, the chance of getting 532 or more heads is 2.3%; in other words, the p-value is 2.3%. 137 This is called one-tailed testing. 138 Alternatively, a statistician can compute the chance of getting 532 or more heads or 468 heads or fewer. 139 The p-value for this example would be 4.6%. 140 This is called two-tailed testing. 141

We discuss these basics of statistical analysis to accent the fact that at the district court level, the parties divided over which test was more appropriate: one-tailed or two-tailed. While agreeing that a p-value of 5% or smaller showed statistical significance for two-tailed testing, they disagreed about the significance level for one-tailed testing. The State argued that a p- value of 2.5% or smaller showed statistical significance for one-tailed testing, while Woodfox argued that 5% or smaller would do.

The district court found it unnecessary to solve these problems.- Under the 40.8% eligible population baseline, the p-value for one-tailed testing was 1.26% and for two-tailed testing was 1.85%. Both p-values were below the threshold required to show statistical significance. We do not find any clear error in the district court finding.

Therefore, the State’s attempt to rebut the prima facie cases using statistics does not persuade. The district court did not err in finding as such.

VIII

The State also renews its' arguments that it rebutted the prima facie case by demonstrating the use of race-neutral criteria in the selection of grand jury forepersons. Such a rebuttal case operates by “showing that permissible racially neutral selection criteria and procedures have produced the monochromatic result.” 142 “[Affirmations of good faith in making individual selections are insufficient to dispel a prima facie case of systematic exclusion.” 143 But the “presumption of discriminatory conduct may be successfully rebutted by testimony of responsible public officials if that testimony establishes the use -of racially neutral selection procedures.” 144

During the time relevant time period, the two judges of the 21st Judicial District appointed grand jury forepersons in West Feliciana Parish: the late Judge Ramshur and Judge Kline.

Judge Kline testified at the federal evidentiary hearing. According to Judge Kline, he would think of someone who would be a good foreperson and would attempt to contact them during the morning of the venire. If he did not know someone on the venire, he sought facts about the person, not opinions, about whether he or she would be good foreperson. Judge Kline explained that various criteria mattered, including character, communication skills, patience, independence, reputation and education. But while education and employment were important, they were not determinative. Instead, Judge Kline sought “basic education” and looked' for employment be *382 cause it “reflected some dependability.” But Judge Kline also stated that he did not want to choose only people with advanced degrees because that would eliminate “a whole body of good folks with good common sense.” Judge Kline also stated that he actively tried to be inclusive, and appointed women and African-American forepersons without as much education as others in the pool but who were “representative of the community.” Finally, Judge Kline clarified that he could only speak to his own selection procedures.

Because Judge Ramshur passed away in 2006, the State presented other officials familiar with his selection process. The State presented Judge George H. Ware, Jr., who was the District Attorney of West Feliciana Parish from 1985 through 1996. Ware testified that he would meet with the judges as they were selecting the foreperson and discuss potential selections. He testified that the question the judges asked him suggested they were seeking information about “community leadership role, responsibility in the community, background, whether or not this person was a gossip.” Occasionally, the judges would ask him questions about the potential foreperson’s job and family. The State also presented Jesse Means, the Assistant District Attorney for the 20th Judicial District from 1985 through 2006. Means testified that while Judge Ramshur never asked him for advice, on one occasion, he advised the Judge not to select a person. • But this testimony was given only under proffer as it was hearsay testimony. In totality, Means testified that he did not give Judge Ramshur “specific advice about specific people” because the Judge did not need it. Finally, the State also presented other witnesses to attest to the race-neutral selection by Judge Ramshur. Much of this testimony is irrelevant or was offered under proffer. As an example, the former clerk of the court in East Feliciana Parish testified as to what she thought Judge Ramshür’s practices were in East Feliciana. 145

This rebuttal evidence is more that affirmations of good faith that discrimination did not occur, but it is not the sort of evidence that rebuts a prima facie case. 146 The State contends that subjective criteria like “character” and “leadership” are acceptable. We need not disagree, although our past pronouncements have created some confusion on this point. 147 But the difficulty is that while Judge Kline was able to articulate race-neutral criteria, there is almost no evidence that Judge Ramshur employed race-neutral criteria, either objective or subjective. What makes matters worse is that the only information the judges received about the people on the grand jury venire *383 were the names, addresses, and in later years, the telephone numbers. As we have noted before, “[t]he presence of identified objective criteria known in advance to the appointing judge would have mitigated the difficulties of the selection system then in place.” 148 As far as we are able to discern, Judge Ramshur mostly selected people known to him without any systematic attempt to obtain information about qualifications. Indeed, Judge Ramshur’s grand jury venire transcripts shows a lack of questions as to qualifications. We also take special note of the fact that of the five African-American grand jury forepersons during the relevant time period, Judge Kline selected four. Thus, Judge Ramshur selected only one. Indeed, as Woodfox points out, Judge Ramshur selected a grand jury foreperson nineteen times during this same period.

The State does suggest a plethora of race-neutral criteria that can account for the disparity, such as employment, education, character, and independence. In support, the State provided a lot of data. First, it compiled a list of grand jury forepersons between 1980 and March 1993 to show that they all shared similar education and employment characteristics. Second, it produced U.S. census data showing educational attainment by race for 1980 and 1990. These data corroborate that African-Americans were less educated than the general population. Third, it produced U.S. census data showing unemployment and lack of participation in the labor force by race for 1980 and 1990. These data also corroborate that African-American were less employed and participated less in the labor force that the general population. Yet the problem with this evidence is that it fails to persuade when considered in light of the fact that there is no evidence Judge Ramshur actually knew about the characteristics when picking the foreperson. 149

The State’s argument that West Feliciana Parish is small and the judges knew all its members is no more than a good faith assertion. Moreover, the State’s assertion that the judges made proactive attempts to include women and minorities fails to convince. Only Judge Kline made such an assertion, and we have no reason to believe that Judge Ramshur made similar attempts. Furthermore, the records reveals that Judge Ramshur in particular passed over equally qualified African-American candidates to appoint white forepersons. Woodfox identified specific African-American venire members and their employment and education, and compared those qualifications to the white forepersons actually selected. For almost every year, Woodfox can point to African-Americans in the grand jury venire that had comparable educational and.employment experience to the selected foreperson. This bolsters our conclusion.

We hold then that the State has not demonstrated reversible error in the district court’s holding that it failed to rebut the prima facie case.

IX

For these reasons, we AFFIRM the district court’s grant of habeas relief; and REMAND for further proceedings consistent with this opinion.