Darryl Shirley v. James Yates

807 F.3d 1090

Opinion

OPINION

REINHARDT, Circuit Judge:

I.Introduction

Darryl Shirley was convicted of the first-degree burglary of an unoccupied residence and the second-degree robbery of a sandwich shop (he took $80 from the cash register). In neither instance was anyone harmed, and no weapons were involved in either offense. Shirley was sentenced to two consecutive 25-years-to-life terms in prison for the burglary and robbery, and also four consecutive five-year sentence enhancements based on prior convictions.

On habeas, Shirley properly raised a number of claims. Because we reverse the district court’s denial of relief on his claim under Batson v. Kentucky, 476 U.S. 79, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986), we need not reach his other claims.

II.The Batson Framework

Batson sets out a three-step burden-shifting framework for evaluating claims of discriminatory peremptory strikes. At Step One, the defendant bears the burden to “produc[e] evidence sufficient to permit the trial judge to draw an inference that discrimination has occurred.” Johnson v. California, 545 U.S. 162, 170, 125 S.Ct. 2410, 162 L.Ed.2d 129 (2005). Once the defendant makes out a prima facie case, at Step Two “the burden shifts to the State to explain adequately the racial exclusion by offering permissible race-neutral justifications for the strikes.” Id. at 168, 125 S.Ct. 2410 (internal quotation marks omitted). Finally, at Step Three, “[i]f a race-neutral explanation is tendered, the trial court must then decide ... whether the opponent of the strike has proved purposeful racial discrimination.” Id. (internal quotation marks omitted).

III.Procedural History

1. State Trial Court

At Shirley’s trial, a 60-person venire was empaneled and sworn. Of that number, five veniremembers were black. (Shirley, too, is black.) Of those five, all but one were removed from the venire— the remaining one, who was among the twelve originally summoned to the box, was seated on the jury. First, J.H. was dismissed by the court for cause, due to illness. Next, L.L. was peremptorily struck by the state. Then, K.A. was dis missed for cause on the parties’ joint motion, because she had a brother with a criminal record and said that she would have trouble sitting in judgment of another person. Finally, the state used another peremptory strike to dismiss R.O. 1 After this strike, Shirley made a Batson motion, claiming that the peremptory strikes of L.L. and R.O. were racially discriminatory. The motion was denied, with the trial judge stating that Shirley had failed to make out a prima facie case:

[L.L.] had a misdemeanor conviction in her background, related to fraud, which may have reflected, and in fact, did reflect on her moral turpitude.

[R.O.] ostensibly appeared to be an acceptable juror. She was young, although she did express an interest in being on the panel.

But except for [R.O.j’s possible improper excusal, I don’t see any pattern of exercise of improper peremptory challenges by the People.

So I And there is no prima facie case or demonstration of an improper exercise of excusal of peremptory challenges against African Americans, especially in light of the fact that we have one original juror, [ ], still on the jury, or at least potential jury, who is one of the original 12 who were summoned in the jury box, who is still present.

2. State Court of Appeal

The California Court of Appeal affirmed the trial court’s Batson ruling, relying on People v. Box, 23 Cal.4th 1153, 1188, 99 Cal.Rptr.2d 69, 5 P.3d 130 (2000). See People v. Shirley, 2007 WL 1302512 at *4 (Cal.Ct.App. May 4, 2007). The Court of Appeal quoted Box stating that “when the record ‘suggests grounds upon which the prosecutor might reasonably have challenged the jurors in question, we affirm.’ ” It concluded that there were two race-neutral reasons for dismissing L.L.: the prior misdemeanor conviction for fraud, and her possible familiarity with the defendant and one of his relatives. It also concluded that R.O.’s “age and corresponding lack of life experience” was a legitimate race-neutral reason for striking her. The state court added that any inference of discrimination with respect to the strike of R.O. was undermined by the fact that three young white veniremembers who “demonstrated a lack of life experience” were also struck. “[T]he record also shows,” the court noted, “that the individuals who were selected to sit on the jury were reasonably intelligent and had a good deal of work and/or life experience.” The state court observed that while another veniremember — one who was seated, as Juror Number 3 — “was a senior at ‘Sac State’- who lived with his parents, he also worked as the manager of a gym, a position that involves decision making.” Finally, it speculated that the prosecutor might have been seeking a “strategically balanced” jury, such that “one young juror on the panel was sufficient.” The state court’s reasons were not based on any representations made by the prosecutor but were inferred by the court from its examination of the jurors’ qualifications as contained in the record.

3. District Court

Shirley filed a federal habeas petition. The district court concluded that the state court had acted contrary to clearly established Supreme Court precedent by finding, on the basis of speculation about possible race-neutral reasons for exercising the challenged strikes, that Shirley failed to raise an inference of discrimination and thereby make out a prima face case. Shirley v. Yates, 2013 WL 394713 at *2 (E.D.Cal. Jan. 30, 2013). On de novo review, the district court concluded that Shirley had satisfied Batson Step One by showing that two out of three eligible black veniremembers were peremptorily struck and that the second, R.O., was similar to a white veniremember who was seated. Because the state court “prematurely cut off the Batson inquiry at Step [One],” the district court conducted a hearing to take evidence regarding the prosecutor’s reasons for exercising the challenged strikes.

A. Evidentiary Hearing

At the evidentiary hearing, the court heard testimony from the deputy district attorney who prosecuted Shirley eight years earlier, Alan Van Stralen. Van Stralen testified that he had a “general approach to jury selection,” which was “well entrenched” prior to Shirley’s trial, Van Stralen’s fifty-fourth. Van Stralen was asked to describe his “criteria for identifying desirable or undesirable jurors:”

Things that I would like to see in a prospective juror would include intelligence, I guess, first of all. And I would determine that through how the person presents themselves in court, the manner in which they speak, the type of employment they’ve held. A college education is definitely a plus as far as that goes.

I like to see jurors who have life experience. That’s quite important to me. And I, in that regard, would look at things such as — well, a person basically who has been around, done some things, who’s been in different situations, met different people.

In that regard I, again, would put some significance on a college education because that does get a person out into the world and new environments and meeting new people and encountering new situations and introduces them to analytical thinking.

I also would look to and place a lot of emphasis on a person’s employment. If a person has employment — first of all I like to have people who are employed and have a good track record of employment. But in particular what I like to see is people who have jobs that would involve a significant amount of responsibility or a significant amount of decision-making.

I also like jurors who, to the extent I can ascertain this, would appear to have a personality that would mesh with other people, somebody who would get along with other people since for a successful prosecution I’m looking for 12 people to be able to agree.

So I like to get people on my jury who are likely to be able to get along with others and do well in the deliberation process.

Actually, I also look kind of overall — ]has this person led a responsible life. And, you know, I might, again, look to the type of employment they’ve had, what their family history is, were they raised in a family, do they have any run-ins themselves with the law. That would not be a favorable aspect' as far as I’m concerned.

And on a similar note, if somebody had a favorable — particularly favorable attitude towards law enforcement, that would be a plus for me because a person like that, I believe, would trust the system enough that if the evidence was there to prove guilt, they would not hesitate to vote for guilt.

Because he listed primarily positive factors, Van Stralen was asked to elaborate on the one “potentially negative” factor he had mentioned. He explained that he considered crimes committed “well into somebody’s adulthood,” or those involving “dishonesty,” such as theft, perjury, or fraud, to be “deal breaker[s].” Asked whether there were “other negative qualifications”; he said, “[p]robably just the opposite of the things that I just went through as far as what I would look for in a juror.” 2

Van Stralen went on to explain that he had “a practice of making contemporaneous notes when selecting a jury” on “small yellow Post-its.” He stated, however, that he “[did] not retain those notes,” suggesting that “[o]ther people might be better at taking notes than I am,” and explaining that he did not “find ... a lot of time to write down detailed notes during jury selection,” but. rather “jot[ted] down just brief little comments or words ... a shorthand.” “I don’t believe those notes would be useful to me years down the road,” Van Stralen testified. 3

Although Van Stralen did recall the facts of Shirley’s case, and did recall picking a jury, he did not recall the reasons he exercised his strikes. In addition, 'he did not keep the notes he made during voir dire, and the prosecution’s case file “was no longer available” — -“[i]t had possibly been destroyed ... after a period of time.”

Based on his review of the voir dire transcript, 4 Van Stralen testified that he struck L.L. for three reasons. First, because L.L. stated that Shirley looked familiar to her, it was possible that she “might have a bias in favor of [him],” or “might be a bit more reluctant than a person who didn’t know or didn’t find the defendant familiar to convict because she might believe that she would see him around again.” Second, L.L. had recently met — and was considering conducting business with — a man who may have been related to Shirley. Third, L.L. had been convicted of a crime as an adult, which “would lead [Van Stralen] to believe she hasn’t led as responsible a life as [he] would hope for [his] prospective jurors.” 5

As for R.O., Van Stralen indicated that he struck her “simply because she lacked life experience”: she was “very young ... three years out of high school,” and worked at a Walgreens pharmacy developing photographs. “She had not gone off to college. Apparently she’s living and working in the same area she grew up in.” Van Stralen explained that he “thinkfs] people who lack that kind of life experience don’t make particularly good jurors. They don’t have a perspective upon which to make sound decisions.”

When asked, for purposes of a comparative juror analysis, about a white venire-member of approximately the same age he allowed to be seated, Van Stralen told the court that the veniremember — Juror Number 3 — was desirable because he was “going to college[,] [which] shows some initiative [and] a certain degree of intelligence generally speaking.” Juror Number 3 was also a manager at a gymnastics facility, which showed Van Stralen that he had “significant responsibilities ... and would be involved in decision-making.” “[H]e had something going for him ... [and] did have life experience ... [and] a perspective that I think would be valuable as a juror.” Van Stralen also noted that he “put some consideration on the fact that” Juror Number 3 had a “favorable view of law enforcement,” reflected by the fact that he twice called the police to report vandalism or burglaries at the elementary school across the street from his house, and that his sister and brother-in-law were employed by police departments.

Van Stralen said that he had a “very, very high degree of confidence” that the reasons he gave in his testimony at the hearing, were his actual reasons for exercising the contested strikes at trial. Critically, however, he was' asked on direct “whether the reasons you dismissed the two jurors ... were based on a specific memory you have or rather were consistent with a standard practice you had in jury selection or something else?” Van Stralen responded:

[B]est way I can answer is after reviewing the transcript, knowing how I approach jury selection, knowing that I have done so in the same fashion for many, many years, including during this period of time, I can say again, without reservation, that those were my reasons because those are reasons that I would dismiss a juror in similar circumstances in any case.

When he reiterated the same on cross, he was asked:

Q. Because that’s your general practice, your general approach?

A. Yes.

Q. And not because you remember [R.O.] or [L.L.]?

A. Correct. I don’t remember....

During a brief oral argument following Van Stralen’s- testimony, Shirley’s counsel argued that Van Stralen “may have had good reasons as he stated, but the real reasons are at the heart of this inquiry. And I don’t think we can find those because he has no recollection of the actual real reasons.” Counsel for the state disputed the legal — but not the factual — assertion, stating that “the evidence before this court is that keeping these two disputed jurors would have been inconsistent with the guidelines that Mr. Van Stralen described as comprising his general approach to selecting a juror. It would have been inconsistent, and that is evidence, circumstantial evidence, of nondiscrimination.” In conducting this evidentiary hearing, the district court satisfied its obligation under Step Two of Batson.

B. Ruling

After the evidentiary hearing, the district court ruled orally. First, the district court determined that the testimony offered by Van Stralen was not sufficient to satisfy the state’s burden of production at Batson Step Two. In so doing, the district court answered a question left open in Paulino v. Harrison (Paulino II), 542 F.3d 692, 701 n. 10 (9th Cir.2008), and at the center of this case: “whether a list of standard considerations, absent affirmative evidence that they were used in the particular case in question, is competent evidence of a prosecutor’s actual reasons for striking certain jurors.” The district court concluded “that you have to have actual evidence of actual reasons,” and that a court cannot “infer it from practices and apply it to a particular case.”

Proceeding on to Batson Step Three, the district judge observed:

[Rjegardless of that finding the court has to go on and make an actual determination whether there was racism involved.

So you’ve then got the presumption that’s created by the prima facie case. You have ... no evidence one way or the other, as I consider it to be evidence, on [S]tep 2. And the question is, based on that, can I infer discrimination.

[I]t is clear that the burden stays on the person ... wishing to challenge the strike....

As to Shirley’s prima facie case, the district court drew a distinction between the number of minority veniremembers struck in Paulino II — there, five out of six strikes were challenged as discriminatory — and the number at issue in Shirley’s case (only two). The district judge then went on to consider whether the reasons the prosecutor gave for striking L.L. “ma[de] sense” to him. He concluded that her removal from the venire “just strikes me as entirely reasonable”; “even putting aside the misdemeanor conviction for what appears to be some sort of dishonesty-related offense,” he said, “I think a prosecutor would be very likely to strike somebody” who might have recognized the defendant and might be doing business with a relative of his. He continued: “And even without the witness’s explanation, once I put it together in my mind, it just seems plausible. And, you know, that’s what he said, and it makes sense.”

The strike of R.O. was, by contrast, “very close.” 6 Rather than assessing whether Van Stralen’s stated reason held up — whether it was the actual reason R.O. was struck — the district judge compared her responses to Juror Number 3 and drew a distinction between them:

She did not, in fact, go to college so that distinguishes her from the other juror who was going to Sacramento State. So they, were about the same age, but she was obviously less educated. I can sort of see prosecutors wanting somebody who has got an education.

On the basis of his conclusion that R.O.’s lack of higher education could have been a good reason to strike her, the district judge denied relief.

IY. Analysis

1. Batson Step One

In Johnson v. California, 545 U.S. 162, 166-67, 169-73, 125 S.Ct. 2410, 162 L.Ed.2d 129 (2005), the Supreme Court held that the California Supreme Court had been wrong to require Batson claimants to show a “strong likelihood” of discrimination at Step One; rather, Johnson reiterated that a defendant makes out a prima facie case if he produces evidence sufficient to support a “reasonable inference” of discrimination.

As the district court correctly concluded, the California Court of Appeal acted contrary to clearly established law when it “based its prima facie analysis on the discredited, pre-Johnson, standard articulated by the California Supreme Court in People v. Box.... ” The state court held that “when the ‘record suggests grounds upon which the prosecutor might reasonably have challenged the jurors in question, we affirm’ ” the trial judge’s ruling that the defendant failed to make out a prima facie case. Based on just such speculation, it found that Shirley had not met his burden at Batson Step One.

We have made clear, however, that Box imposes too high a burden, and that state court decisions applying it do not warrant deference under AEDPA, because they are “contrary to ... clearly established Federal law.” 28 U.S.C. § 2254(d)(1). See Johnson v. Finn, 665 F.3d 1063, 1069 (9th Cir.2011) (“[T]he existence of ‘grounds upon which a prosecutor could reasonably have premised a challenge,’ does not suffice to defeat an inference of racial bias at the first step of the Batson framework.”). Hence, it was appropriate for the district court to determine de novo whether Shirley had raised an inference of racial bias. Id. at 1069-70.

We likewise agree with the district court that, contrary to the state court’s conclusion, Shirley did raise an inference of discrimination more than sufficient to meet his “minimal” burden at Bat-son Step One. Id. at 1071. The fact that a prosecutor peremptorily strikes all or most veniremembers of the defendant’s race— as was the case here — is often sufficient on its own to make a prima facie case at Step One. See Paulino v. Castro (Paulino I), 371 F.3d 1083, 1091 (9th Cir.2004) (“[A] defendant can make a prima facie showing based on statistical disparities alone.”). In this case, two-thirds of the black veniremembers not removed for cause were struck by the prosecutor. We have found an inference of discrimination in cases where smaller percentages of minority veniremembers were peremptorily struck. Fernandez v. Roe, 286 F.3d 1073, 1078 (9th Cir.2002) (56%); Turner v. Marshall (Turner I), 63 F.3d 807, 812 (9th Cir.1995) (56%), overruled on other grounds by Tolbert v. Page, 182 F.3d 677, 685 (9th Cir.1999) (en banc); accord United States v. Alvarado, 923 F.2d 253, 255 (2d Cir.1991) (57%). 7

That one black juror was eventually seated does weigh against an inference of discrimination, but “only nominally” so. Montiel v. City of L.A., 2 F.3d 335, 340 (9th Cir.1993). “In denying a Batson motion, ... a trial court may not rely solely on the fact that some African-Americans remain on the jury.” Turner I, 63 F.3d at 811-14 (finding a prima facie case of discrimination despite the presence of four black jurors). In Fernandez, we observed that “the lone Hispanic juror’s presence on the jury here is less helpful ... in light of the trial judge’s explicit warning to the prosecutor that any additional challenges against Hispanics would trigger a prima facie finding of discrimination.” 286 F.3d at 1079. The same logic applies to the case before us: the trial judge denied Shirley’s Batson motion, made after two of the three remaining black veniremembers had been struck, “in light of the fact that we have one original [black] juror, [ ], still on the jury, or at least potential jury....” The trial judge found no prima facie showing “at this time of a pattern ” 8 of discriminatory strikes. While this oral ruling was slightly more subtle than the warning in Fernandez, its obvious effect was to put Van Stralen on notice: if he struck the remaining black veniremember, he would be required to give reasons for all three strikes.

Additionally, Shirley’s prima facie case is supported by a comparison 9 between one of the black veniremembers who was struck, R.O., and a white veniremember who was seated, Juror Number 3. Both were in their early twenties and lived at home with their parents. There was no readily apparent reason to strike R.O. — as the district court noted, she “said she was eager to be a juror and would follow the law faithfully [and] also indicated that she had experience making ‘tough calls.’ ” While Juror Number 3 had attended but had not (yet) graduated from college, and had a job that may have involved somewhat more responsibility, they were certainly similar enough — apart from race — to help support an inference of discrimination at Batson Step One. 10

We agree with the district court that Shirley’s Step One showing was “sufficient to permit the trial judge to draw an inference that discrimination ... occurred.” Johnson, 545 U.S. at 170, 125 S.Ct. 2410.

2. Batson Step Two

Just as in Paulino II, in Shirley’s case the state court applied the wrong legal framework and erroneously ended its analysis after Step One; it never reached Steps Two and Three of Shirley’s Batson claim. As a result, the prosecutor was not afforded the opportunity to provide his actual reasons for striking the veniremembers in question, and there could be no reasoning or conclusion as to this critical question to which we could defer under AEDPA. Paulino II, 542 F.3d at 698-99 n. 5; see also Turner v. Marshall (Turner II), 121 F.3d 1248, 1254 n. 2 (9th Cir.1997) (same). Hence, the decision we review here is that of the district court, which conducted the inquiry that must be held at Step Two of Batson.

The district court properly ordered an evidentiary hearing to take testimony from Van Stralen as to his reasons for the strikes of L.L. and R.O. Because this evidence — necessary to adjudicate Steps Two and Three of Shirley’s Batson claim — was absent from the state court record due not to any lack of diligence on Shirley’s part but rather to the state court’s error, 28 U.S.C. § 2254(e)(2) did not bar an eviden-tiary hearing. See Williams v. Taylor, 529 U.S. 420, 432, 120 S.Ct. 1479, 146 L.Ed.2d 435 (2000). The bar to evidentia-ry hearings established by Cullen v. Pinholster, 563 U.S. 170, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011), was also inapplicable because the state court’s decision is not entitled to deference under 28 U.S.C. § 2254(d)(1). Finn, 665 F.3d at 1069 n. 1.

“[S]tep Two is an opportunity for the prosecution to explain the real reason for [its] actions.” Yee v. Duncan, 463 F.3d 893, 899 (9th Cir.2006). As we explained in Crittenden v. Ayers (Crittenden I), 624 F.3d 943, 957-58 (9th Cir.2010):

In the usual case, the Batson analysis takes place during or shortly after jury selection. In those cases, the prosecutor offers a contemporaneous explanation for the strike at step two. Where time has passed since the jury selection, the prosecutor may offer an explanation based on his present recollection of his reasons for striking the juror. Where, as here, time has passed and the prosecutor no longer has a present recollection of his or her reasons for striking the juror, the state may offer an explanation based on circumstantial evidence.

What counts as competent circumstantial evidence of the prosecutor’s reasons for exercising peremptory strikes is an open question: “Our circuit has not addressed whether a list of standard considerations, absent affirmative evidence that they were used in the particular case in question, is competent evidence of a prosecutor’s actual reasons for striking certain jurors.” Paulino II, 542 F.3d at 701-02 n. 10. Because this question straddles the fairly muddy line between Bat-son’s Steps Two and Three, we address it in two parts. First, we consider whether such evidence is sufficient to meet the state’s burden of production at Batson Step Two. 11 In the next section, we address a distinct question: how much weight such evidence should be given in assessing whether the defendant has car ried his burden of persuasion at Step Three.

At Step Two, the state must both (1) assert that specific, race-neutral reasons were the actual reasons for the challenged strikes, and (2) offer some evidence which, if credible, would support the conclusion that those reasons were the actual reasons for the strikes. 12 Id. at 699 (“Batson’s step two requires evidence of the prosecutor’s actual reasons for exercising her peremptory challenges.”).

This is a burden of production, not persuasion. In Purkett v. Elem, the Supreme Court made clear that the state’s race-neutral reason need not be “persuasive, or even plausible,” in order to suffice at this stage. 514 U.S. 765, 767, 115 S.Ct. 1769, 131 L.Ed.2d 834 (1995) (per curiam). Thus, whether the evidence in support of the reason is credible or even whether it “makes sense” is to be determined at Step Three. Id. at 768-69, 115 S.Ct. 1769; see also Crittenden I, 624 F.3d at 958 (“[R]egardless of how the state offers its race-neutral justification, it is not the task of the district court at step two to assess the truth of the explanation.”).

In this respect, it is necessary to bear in mind that the state must offer some evidence of “a reason that does not deny equal protection.” Purkett, 514 U.S. at 769, 115 S.Ct. 1769. In Johnson, the Supreme Court affirmed that for the purposes of Batson, “[i]t does not matter that the prosecutor might have had good reasons ...; what matters is the real reason [the veniremembers] were stricken.” 545 U.S. at 172, 125 S.Ct. 2410 (quoting Paulino I, 371 F.3d at 1090 (alterations omitted)). An equal protection violation has occurred if a black veniremember was actually struck for a race-related reason, even if he could have been struck for a race-neutral reason. Hence, the state must offer evidence at Step Two which is probative of the actual reason that a prosecutor exercised the strike at issue. Evidence that a good reason for a strike existed is insufficient in itself at Step Two.

Our cases reflect that voir dire transcripts may be relevant to the Step Two inquiry in three different ways. 13 First, a voir dire transcript (like contemporaneous notes) may be used to refresh a prosecutor’s recollection; upon reviewing it, he may recall the actual reasons for his strikes. This sort of refreshed recollection is clearly sufficient for the purposes of Step Two. See, e.g., Turner II, 121 F.3d at 1251. Second, as in the case before us, a prosecutor may review a voir dire transcript and still remain unable to remember his reasons for striking the particular veniremembers at issue. He may then infer reasons from the transcript and assert them, but that assertion must be supported by circumstantial evidence that tends to show that the asserted reasons were in fact the actual reasons for the strikes. Such circumstantial evidence may consist of the prosecutor’s jury selection notes or, as is the case here, his typical or usual practices or approach to jury selection. 14 Third, as in Paulino II, a prosecutor may offer “mere speculation drawn from her reading of the voir dire transcript.” 15 542 F.3d at 696. Such reasons have not been “ ‘reconstructed,’ as that term is used in Batson cases; they [have been] constructed out of whole cloth ... [and are] nothing more than guess[es]-” Id. at 700. As we held, “[n]o authority supports the State’s claim that pure speculation qualifies as. circumstantial evidence of the prosecutor’s actual reasons, simply because it was the prosecutor herself who offered the speculation during the course of an evidentiary hearing.” Id. at 701. Although the sort of speculation contained in the third category is insufficient to meet the state’s burden of production at Step Two, the court must in any event proceed to Step Three, because the “failure to provide an explanation for exercising a strike does not relieve the trial court of its responsibility [at step three] to make the ultimate determination of whether there has been purposeful discrimination.” Paulino II, 542 F.3d at 702 (quoting Yee, 463 F.3d at 901). However, when the state fails to meet its Step Two burden of production, “there is no race-neutral evidence to weigh,” and the defendant will prevail at Step Three of his Bat-son challenge in almost all such cases. Paulino II, 542 F.3d at 703.

We hold that in the case before us the circumstantial evidence falls in the second of the three categories discussed above, and that the state has met its burden of production at Batson Step Two. Where the'prosecutor who made the strike has no memory (refreshed or otherwise) of his reasons for doing so, but asserts on the basis of his review of the transcript that a race-neutral reason was in fact the reason for the strike, and then supports that assertion with his testimony that his general jury selection approach would have motivated him to strike the veniremember for that reason, such circumstantial evidence is sufficient for the limited purpose of Step Two. Van Stralen’s testimony that he employed a certain jury selection approach, and that employment of that approach would have motivated the strikes he exercised at Shirley’s trial, is relevant (to whatever degree) to the question whether his asserted reasons for striking L.L. and R.O. were his actual reasons for doing so. 16 The inferential link resulting from evidence of a particular practice or approach may be fairly strong or vanishingly weak; still, its strength or weakness is a question for Step Three, see Purkett, 514 U.S. at 767-69, 115 S.Ct. 1769; Crittenden I, 624 F.3d at 958, at which Step the answer may prove dispositive of the ultimate inquiry.

There are some limits to this 'proposition, however. In Paulino II— which falls in the third category outlined above — the state claimed that the prosecutor had “testified as to her ‘general principles’ of jury selection” 17 ; but, we concluded that because she admitted that she “was not sure which of those general principles she considered in selecting the [] jury,” her “statement that she generally sought fair jurors but did not really know what she considered in this particular trial” amounted to “nothing more than a general assertion that her actions were not racially motivated.” 18 Id. The prosecutor supported her guesswork with inferences anyone might (or might not) draw from the voir dire transcript, but, in failing to relate her particular considerations to the strikes she exercised, offered no evidence — direct or circumstantial — that would lend support to an inference that the speculative reasons were her actual reasons. 19

Our holding is consistent with our analysis in Crittenden I. We held there that the district court had not erred in accepting evidence of the prosecutor’s general approach at Step Two. 624 F.3d at 952-53, 958. However, Crittenden I did not settle the question left open in Paulino II, because the prosecutor also offered contemporaneous notes that strongly supported the contention that the asserted reasons were the actual reasons. Id.

In sum, Van Stralen’s testimony was sufficient to meet the state’s burden of production at Step Two because he credibly testified to a jury selection approach— persuasive or unpersuasive, reasonable or unreasonable, plausible or implausible, see Purkett, 514 U.S. at 767, 115 S.Ct. 1769-that supported his asserted reasons for his challenged strikes. Nothing more was required at this stage of the proceeding.

3. Batson Step Three

At Step Three, we hold that the district judge clearly erred in denying Shirley’s Batson claim with respect to R.O. largely on the basis of a comparative juror analysis with a young white veniremember who was seated. 20 The ultimate question in Batson cases is “whether the defendant has proven purposeful discrimination[, which] is a question of fact that we review for clear error.” Paulino II, 542 F.3d at 699; see also Crittenden v. Chappell (Crit-tenden II), 804 F.3d 998, 1005-06 (9th Cir.2015). Although the district judge had just decided that Van Stralen’s testimony was not probative of the actual reason for his strikes, he then concluded that in light of the comparison a challenge to R.O. could have been reasonable and that therefore no Batson violation had occurred, stating that he could “sort of see prosecutors wanting somebody who has got an education.” The district judge did not, however, find that Van Stralen actually struck R.O. for that reason. 21 Again, “[i]t does not matter that the prosecutor might have had good reasons ...; what matters is the real reason [the veniremembers] were stricken.” Johnson, 545 U.S. at 172, 125 S.Ct. 2410 (quoting Paulino I, 371 F.3d at 1090 (alterations omitted)). Because the district court failed to assess whether the circumstantial evidence Van Stralen offered supported the conclusion that his asserted, race-neutral reason for striking R.O. was actually his reason for doing so and because the court failed to make a finding as to the actual reason for the challenge, the district court committed clear error. We now proceed to examine the record to “determine whether the actual reason for the strike violated the defendant’s equal protection rights.” Yee, 463 F.3d at 899.

“[A] defendant opposing a peremptory challenge bears the ultimate burden of proving the challenge was improper[,]” id. at 895; he must carry this burden by a preponderance of the evidence, Paulino II, 542 F.3d at 702. Along with the statistical evidence showing that black veniremembers were disproportionately struck, we weigh the persuasiveness of the state’s evidence as to the prosecutor’s proffered, race-neutral reasons, and any other relevant factors. 22

At Step Three, if a trial judge determines that a prosecutor’s explanation is credible, that may well be the end of the inquiry, but only if the prosecutor has asserted that specific race-neutral reasons were in fact his actual reasons for the strike. In this case, when the district court determined that Van Stralen was credible, all the credited testimony could establish was that the jury selection approach to which Van Stralen testified was “in fact” his general approach. It remained to be determined how much support this circumstantial evidence actually provided for the conclusion that the race-neutral reasons he offered in fact motivated the particular strike.

At Step Three, we hold that the district court clearly erred in finding that Shirley had not met his burden of showing discrimination by a preponderance of the evidence with respect to the strike of R.O., a question the district judge repeatedly said was “very close.” We so hold not because we disbelieve the testimony offered by Van Stralen, but because, although Van Stralen’s approach to jury selection falls within a category of circumstantial evidence sufficient to meet the burden of production at Step Two, the particular showing the state made here provides too little support for its contention that Van Stralen actually struck R.O. for the reason posited.

A.

Circumstantial evidence of a prosecutor’s general jury selection approach is adequate at Step Two, but it deserves little if any weight at Step Three when it describes only a vague general preference as to the type of jurors the prosecutor finds most desirable — e.g., “things that I would like to see in a prospective juror” — rather than the types of veniremembers the prosecutor customarily strikes. Nor does the fact that the absence of such ideal qualifications is á negative factor tell us much about a prosecutor’s practices with respect to actually making strikes.

Circumstantial evidence regarding a prosecutor’s approach to jury selection in cases in which the prosecutor does not recall the specific challenge or the reasons therefor suffers generally from several weaknesses. First, the evidence is not contemporaneous — i.e., generated during jury selection. 23 The Supreme Court has emphasized the importance of the contemporaneous airing of justifications, which helps to avoid the “risks of imprecision and distortion from the passage of time.” Miller-El v. Cockrell (Miller-El I), 537 U.S. 322, 342-43, 123 S.Ct. 1029, 154 L.Ed.2d 931 (2003). A prosecutor asked to produce reasons on the spot will not have time to review a transcript to select a reason that is supported by the record (and was consistently applied). Van Stralen testified about his approach eight years after the fact, and after reviewing the reasons the state court had decided were warranted by the record and would have been good enough. Second, prosecutorial approach evidence is not specific to the particular trial and particular strike at issue. Batson itself makes clear that lest the promise of equal protection become a “vain and illusory requirement,” a prosecutor must offer not “general assertions” but rather a “neutral explanation related to the particular case to be tried.” 476 U.S. at 98, 106 S.Ct. 1712 (emphasis added). Van Stralen was able to tie his approach to Shirley’s trial only by stating that he “approaeh[ed] jury selection ... in the same fashion for many, many years, including during this period of time.”

Third, even if the circumstantial evidence credibly demonstrates that a general jury selection approach was actually employed at a given trial, it is possible that another extraneous factor — race or something else — motivated a particular strike, at least in substantial part. As we explained in Crittenden I, “the proper analysis at Batson’s step three is whether the peremptory strike was ‘motivated in substantial part’ by race. If it was so motivated, the petition is to be granted regardless of whether the strike would have issued if race had played no role.” 624 F.3d at 958-59 (quoting Cook v. LaMarque, 593 F.3d 810, 815 (9th Cir.2010)). When a prosecutor testifies that he recalls striking a black veniremember because he prefers jurors who are older and this veniremember was young, he is implicitly attesting that he recalls striking the veniremember for that reason and not for others such as race. But when the prosecutor does not remember the strike, evidence that he prefers older jurors tells us only that this consideration may have played some role in his selection decision, and says next to nothing about whether other, illicit, factors also motivated the strike.

As this case illustrates, approaches to jury selection fall along a spectrum: at one end are practices of regularly striking veniremembers who have a certain objectively-defined characteristic; at the other end are vague preferences for or against jurors with more or less of a particular subjective attribute. For example, a prosecutor may testify that if a veniremember has not graduated from high school, he will usually strike him on that basis alone — a regular practice — or he may testify that he likes jurors who “got a good education” — a mere preference for an experience only vaguely defined. Weaker still is evidence of a general preference for jurors with an undefined characteristic or set of experiences, such as that which the prosecutor described here — “life experience.”

Because evidence of a prosecutor’s approach to jury selection logically supports his asserted reason for a strike only to the extent we can conclude that it motivated a particular strike, only a practice of regularly striking veniremembers with specific traits, characteristics, or similar disqualifying features or attributes, can ordinarily provide persuasive evidence that an asserted reason for a strike was the actual reason for its exercise. 24 If a prosecutor simply says that he generally prefers educated jurors or jurors with adequate job experience but does not remember the strike itself, it will be impossible to ascertain — as we must — whether this preference played any role in the decision to strike a particular veniremember and, even if it played some role, whether it was a determinative one. Nor can we determine with any degree of confidence whether the strike would have been exercised if the veniremember had been white instead of black. In short, a general approach to what constitutes preferred jurors, standing alone, affords us no way to determine with respect to a particular strike the importance of the fact that the veniremember had not attended college and instead proceeded straight to the business of earning a living, or that he had worked for three years in a retail store as compared to five as an auto mechanic. Indeed, a general preference cannot tell us in what circumstances a prosecutor would strike a venire-member for falling short of his ideal.

The value of the juror-preference type of evidence is at its nadir when the prosecutor expresses his preference in terms as startlingly vague and subjective as Van Stralen’s stated desire for jurors “who ha[ve] been around, done some things, ... been in different situations, met different people.” Such a preference tells us only whether the juror meets all the prosecutor’s highest aspirations, not how to assess the value of the “things” a given venireper-son has “done,” nor, most important, in what circumstances the absence of these “things” would cause the prosecutor to exercise a strike. Van Stralen’s testimony that he is certain — though he cannot recall — that a particular veniremember lacked such “things,” and that he struck her for that reason, constitutes an inference he drew that amounts to little more than “rank speculation” — an inference that is not supported by probative evidence, circumstantial or otherwise. Paulino II, 542 F.3d at 699.

The sorts of characteristics likely to warrant a regular practice of striking veniremembers — and to serve as persuasive circumstantial evidence — are those least likely to be pretextual; such practices will ordinarily be concrete and motivated by a patently legitimate purpose. Regular practices of striking venire-members are likely to apply to clear-cut cases that pose little risk of post-hoc rationalization. 25 Moreover, it is likely that it will be evident from a review of the transcript whether such practices are consistently applied. By contrast, it will often be impossible to tell whether and to what extent a general preference motivated a particular strike, especially a vague preference, such as one for jurors with “life experience.” 26 This difficulty in telling what role, if any, a general preference played in the strike may frequently be resolved by the prosecutor’s contemporaneous notes.

With the distinction in mind between a practice of striking veniremembers with particular specific attributes and a preference for jurors with general types of experiences, we turn to evaluating the circumstantial evidence Van Stralen offered in the district court.

B.

Van Stralen gave two reasons for striking L.L.: her criminal conviction and her possible acquaintance with the defendant and a relative of his. He testified that in striking L.L., he followed a regular practice because an adult criminal conviction was a “deal breaker.” The transcript shows that L.L. was convicted of a crime as an adult — a clear and specific factor on the basis of which Van Stralen consistently exercised strikes. Hence, the circumstantial evidence of his regular practice provided significant support for the conclusion that Van Stralen struck L.L. because of her conviction. Although we need not decide here the validity of L.L.’s strike, it provides a clear contrast to the strike of R.O.

The preference Van Stralen expressed, for “jurors who have life experience ... well, a person basically who has been around, done some things, who’s been in different situations, met different people,” is far from describing a regular practice with regard to strikes and is, to the contrary, extremely vague. That Van Stralen “like[s] to see jurors who have life experience” cannot in itself support the conclusion that he struck R.O. for that reason. It is far from evident from the transcript that R.O. had so little life experience that this preference was a significant, much less determinative, factor in Van Stralen’s decision to strike her 27 ‘ and he did not testify that he had any such practice. Instead, Van Stralen suggested that a lack of life experience was a negative factor in his overall review of jurors’ qualifications. This contrasts with his typical practice of striking jurors whose adult criminal convictions he considered deal breakers. Because Van Stralen’s testimony established at best that he may have considered life experience in deciding whether to strike R.O., and not that he actually based his decision to strike her on that ground, his testimony provides little support for his assertion (which was based only on an inference) that this was his actual reason for striking R.O.

C.

The comparative juror analysis does not aid the state’s case; actually, the comparison between R.O. and Juror Number 3, on which the district court so heavily relied, cuts in Shirley’s favor. The purpose of comparative juror analysis is largely to test for consistency. If a prosecutor states that his reason for striking a venire-member was that she had. been convicted of a crime, and that he has a practice of striking all veniremembers with criminal convictions, it is easy to determine whether any of the jurors permitted to serve also had criminal records. When, however, the prosecutor merely states that he believes he struck a veniremember because he has a preference for jurors with “life experience,” and that upon his review of the transcript it looks to him like she had less of it than another juror he permitted to serve, the comparison is less probative.

The comparative juror analysis here shows that a white juror with a very similar level of “life experience” was seated. Van Stralen explained his reason for striking R.O. by identifying indications in the transcript of her limited “life experience,” focusing largely on her youth and also emphasizing that she had not left home. Juror Number 3 was approximately the same age and likewise had not left home. Although Van Stralen stated that Juror Number 3 was enrolled in college at “Sac State” and worked as a gym manager while R.O. had not taken college courses and worked as a photography technician at a pharmacy, these differences were, for purposes of Batson, minor. 28 Moreover, R.O. explained that she was enthusiastic about serving and would readily follow the evidence, while Juror Number 3 said that he preferred not to serve and did not have the ability to follow the trial with his full attention. As in Miller-El II, there were “strong similarities as well as some differences.” There will of course always be at least some differences because even very similar “potential jurors are not products of a set of cookie cutters.” 545 U.S. at 247 & n. 6, 125 S.Ct. 2317. Here, the differences were slight and actually favored R.O. They do not support the state’s claims.

V. Conclusion

Our opinion addresses a narrow set of Batson cases in which the prosecutor cannot actually remember the reason why he struck the veniremembers. In such cases, we hold that if a prosecutor testifies both to his general jury selection approach and that he is confident one of these race-neutral preferences was the actual reason for the strike, this is sufficient circumstantial evidence to satisfy Batson Step Two. Nevertheless, we also hold that this evidence alone will seldom be enough at Step Three to overcome a prima facie case unless the prosecutor has a regular practice of striking veniremembers who possess an objective characteristic that may be clearly defined. That a veniremember (allegedly) lacks a certain je ne sais quoi that the prosecutor prefers is simply not enough. Here, the district court incorrectly found that Van Stralen had not met the state’s burden of production at Step Two. More significantly, the district court clearly erred in denying Shirley’s claim at Step Three on the basis of a juror comparison and its view that the reason Van Stralen proffered could have been a good reason for striking R.O. The district judge did not determine whether Van Stralen had offered circumstantial evidence sufficient to support the inference that he actually struck R.O. for the reason proffered.

Shirley’s prima facie evidence of discrimination was met with only a lengthy statement that Van Stralen “liked to see jurors who have life experience.” His vague, general preference — as opposed to a regular practice of striking venire-members for a specific reason — constituted at most an inclination towards jurors with highly indefinite attributes or qualities. A vague approach to jury selection may constitute sufficient circumstantial evidence for purposes of Step Two, but, in a case in which the prosecutor does not recall his actual reason for striking the juror in question, it provides little or no probative support for a conclusion at Step Three that he struck her for the reason he proffered. Nor does a comparative juror analysis help the state’s case here. Thus, Shirley’s evidence was sufficient to carry his burden of showing that the strike of R.O. was motivated in substantial part by race.

We therefore reverse the decision of the district court and remand with instructions to grant the writ unless the state elects to retry Shirley within a reasonable amount of time.

REVERSED and REMANDED.

1 . The prosecutor was afforded 20 peremptory strikes, of which he used only 10. By the time Shirley made his Batson motion, the prosecutor had exercised nine strikes, to dismiss L.L., R.O., and seven other venire-members (whose race is not apparent from the record, but whose dismissals were not challenged). The prosecutor subsequently used one additional strike.

2 . The full question to Van Stralen was: “And besides the criminal history of a prospective juror, are there other negative qualifications or deal breakers that you might seek to avoid in a juror?” Van Stralen's answer, however, appears to describe only negative qualifications — not deal breakers. He discussed “things I’d like to avoid,” noting they were mostly just the absence of things he would-look for in an ideal juror, for example, "somebody [who] didn't appear tq be particularly intelligent, [or] somebody [who] didn't have a lot of life experience.” By contrast, in response to a question moments later, he testified that a veniremember having committed a "dishonesty” crime or a crime well into adulthood was a "deal breaker” in that he would not accept that person on the jury under any circumstance.

3 . He also stated that he did not — and was not allowed to — retain juror questionnaires.

4 .On cross-examination, Van Stralen revealed that he had also reviewed the state court decision and the district court order finding the state court's Step One determination to be unreasonable and ordering a Step Two hearing. The district court found his testimony at that hearing to be "entirely credible ... very professional ... [and] very forthright,” and we defer to that determination because it is not clearly erroneous. Fed.R.Civ.P. 52(a)(6). However, Van Stralen articulated his purported reasons for striking L.L. and R.O. in terms — literally—that are remarkably similar to those previously hypothesized, without the benefit of his testimony, by the state court and discussed in the district court order. Both the state court and Van Stralen said that R.O. had too little "life experience,” and emphasized that while her job — developing photographs at a drugstore — apparently did not involve "decision making,” the young white veniremember's job — managing a gym — did.

5 . Until counsel for the state refreshed his recollection in the midst of his testimony, Van Stralen could not recall for sure whether he had questioned L.L. about her conviction. He “did not see that in [his] review of the transcript” until he was directed to the relevant pages of the transcript: "I did question her. And that was actually kind of what I was trying to relate a moment ago. I think I may have questioned her, but I don’t know that I did. And the transcript doesn’t reflect that I asked her the nature of the conviction.”

6 . The district judge called it “very close” four times in less than one page of hearing transcript, and noted that "the Court of Appeals might well disagree with me.”

7 . The denominators in these cases — that is, the total numbers of minorities in the veni-res — were larger than three. Wade v. Terhune, 202 F.3d 1190, 1198 (9th Cir.2000), does suggest, quite logically, that larger denominators will lead to stronger inferences of discrimination, and cautions that when a Bat-son challenge is made after the first minority to be called into the jury box is peremptorily struck but well before jury selection concludes, it would be erroneous to find that a prima facie case had been made merely because at the time of the challenge the prosecutor had struck 100% of minority venire-members. In contrast to Wade, Shirley has challenged both of the black veniremembers ■Van Stralen struck, not only one of them. Additionally, in Shirley’s case, peremptory strikes were used at a disproportionate rate against black veniremembers.

8 . Emphasis added.

9 . Because Batson requires us to consider the relevant circumstances surrounding a challenged peremptory strike, 476 U.S. at 96-97, 106 S.Ct. 1712, comparative juror analysis is "called for on appeal even when the trial court ruled that the defendant failed to make a prima facie showing at the first step of the Batson analysis.” Boyd v. Newland, 467 F.3d 1139, 1149 (9th Cir.2006); see also Crittenden v. Ayers (Crittenden I), 624 F.3d 943, 956 (9th Cir.2010) (concluding that the defendant made a prima facie showing at Batson Step One based in part on a comparative juror analysis).

10 . Although Van Stralen’s asserted race-neutral reasons for striking R.O. (i.e., his testimony at Step Two) would inform a comparative juror analysis at Step Three, by specifying that he considered relevant particular distinctions between R.O. and Juror Number 3, the distinctions are irrelevant at the (precedent) Step One. See Fernandez, 286 F.3d at 1079 ("[W]e should not even consider the prosecutor’s unsubstantiated explanations at the stage of determining whether a prima facie case exists.”).

Strikingly, the trial court itself appears to have recognized as much, as it described the strike of R.O. as "possibl[y] improper.” It failed to find a prima facie case of discrimination, however, because it erroneously believed that Shirley was required to show a "pattern of exercise of improper peremptory challenges. ...” Cf. Fernandez, 286 F.3d at 1078 ("A pattern of exclusionary strikes is not necessary for finding an inference of discrimination. See United States v. Vasquez-Lopez, 22 F.3d 900, 902 (9th Cir.1994) ('[T]he Constitution forbids striking even a single prospective juror for a discriminatory purpose.’)”).

11 . "At Batson's second step, the question whether the state has offered a race-neutral reason is a question of law that we review de novo.” Paulino II, 542 F.3d at 699 (citation and internal quotation marks omitted).

12 . While a bald assertion by the state is inadequate, an assertion by a prosecutor that he remembers striking a veniremember for a particular reason is sufficient to meet the burden of production at Step Two, because the prosecutor’s memory-based testimony is direct evidence which, if believed, supports a finding that he actually exercised the strike for the reason articulated. What the state must in any event produce at' Step Two is some evidence regarding the actual reasons for the prosecutor’s action during voir dire, circumstantial or otherwise.

13 . Of course, even in the ordinary case in which a prosecutor gives his non-discriminatory reasons contemporaneously, or remembers them'' at a later hearing, the transcript from voir dire is virtually always important to the Step Three determination.

14 . It is also possible that a transcript itself could contain circumstantial evidence of the actual reasons for peremptory strikes. For example, a prosecutor might have stated, on the record at trial, a reason for striking a particular veniremember, such as in the context of an unsuccessful strike for cause. As we acknowledged in Paulino II, 542 F.3d at 700 n. 8, it is also conceivable that a prosecutor's questions at voir dire might be so sharply focused on a particular issue that the transcript could offer strong support for the conclusion that this was the actual basis for a strike. In this case, however, the voir dire transcript is not so illuminating.

15 . The prosecutor testified that she was "on the same page” as the judge and the defense attorney — “all each could do was comment on the transcript.” 542 F.3d at 696.

16 .Another category of circumstantial evidence not available in this case — contemporaneous notes — may provide far stronger evidence that an asserted reason in fact motivated the strike of a long-since-forgotten veniremember. Notes made during jury selection reflect the prosecutor’s state of mind at the time. Although we may rely on his later testimony (including any discussion of his general jury selection approach) as an interpretive aid, the notes are likely to reveal actual concerns the prosecutor had at trial. Prosecutors who do not retain notes from- voir dire' run the risk that, as here, they will not be able to produce circumstantial evidence of their actual reasons for exercising a strike.

17 . The prosecutor in Paulino II did testify that she "generally ... like[d] jurors who have some good life experience” as well as "older jurors” and “people who would be, of course, fair to all sides.” 542 F.3d at 701. However, she, unlike Van Stralen, did not assert a belief that these considerations motivated her strikes, but rather "offered hypothetical race-neutral reasons” without stating that her jury selection approach was responsible for her decision to exercise the strikes of specific veniremembers. Id. at 696.

18 . Just as protestations of colorblindness are not specific enough to be credited as justification for peremptory strike1s, Batson, 476 U.S. at 98 & n. 20, 106 S.Ct. 1712, testimony as to a general approach of selecting jurors without attention to race would certainly be inadequate to meet the state’s burden at Step Two.

19 . In Paulino II, we cited with approval the Eleventh Circuit’s decision in Bui v. Haley, 321 F.3d 1304 (11th Cir.2003). Paulino II, 542 F.3d at 700-01. Bui involved testimony from an assistant prosecutor who had observed jury selection at trial but had not conferred — either contemporaneously or after the fact — with the lead prosecutor who conducted it about his reasons for exercising the strikes. 321 F.3d at 1310-11. Based on her review of the voir dire transcript and her own notes from trial, the assistant prosecutor offered "representations and argument” as to the reasons for all but one of the challenged strikes. Id. at 1315. The court held, however, that although the assistant prosecutor had asserted specific reasons, .her evidence was not probative of "the reasons [the lead prosecutor] had in mind when he made the strikes.” Id. By contrast, Van Stralen’s testimony regarding his general approach, which he says he applied at trial, is probative (how strongly remains to be determined) of his state of mind at trial.

20 . We assume for purposes of this opinion that the district court was correct to conclude that Shirley failed to carry his burden of showing that Van Stralen’s strike of L.L. was discriminatory. In light of our decision with respect to R.O., there is no need to decide that question. A Batson violation with respect to one veniremember is sufficient to require reversal. See Vasquez-Lopez, 22 F.3d at 902.

21 . Had we agreed with the district court's conclusion regarding the question left open in Paulino II and its related ruling that the state had failed to meet its burden at Step Two, we would have held that the district court erred when it stated at Step Three that there was "no evidence one way or the other, as I consider it to be evidence, on step 2.” This is an incorrect statement of the law — a failure to produce competent evidence at Step Two, whether due to a failure of memory or otherwise, is affirmative evidence of discrimination. See Paulino II, 542 F.3d at 703 ("Where the state fails to meet its burden of production, the evidence before the district court at step three — the prima facie showing plus the evidence of discrimination drawn from the state's failure to produce a reason — will establish purposeful discrimination by a preponderance of the evidence in most cases. Indeed, in such cases, there is no race-neutral evidence to weigh.”).

22 . The only factor the state identifies not otherwise addressed here is that Van Stralen's questioning of the veniremembers . whose strikes are at issue was not desultory. True— but while such questioning (or differential questioning) of minority veniremembers would offer significant support for a finding of discrimination, its absence — as in most Batson cases granting relief — is not especially noteworthy.

23 . We do not consider here the weight to be accorded to a prosecutor's jury selection notes, but limit our analysis to cases involving jury selection practices or approaches.

24 . Accord Polk v. Dixie Ins. Co., 972 F.2d 83, 84-85 (5th Cir.1992) (per curiam) (crediting testimony that a particular factor was the "turning factor in every [jury selection] decision ... in every case”).

25 . In Green v. Travis, then-Judge Sotomayor considered a case in which the prosecutor "had little to no independent recollection of the characteristics or comments of any of the venirepersons” and had no notes regarding one of the five challenged strikes, but explained that she had a practice of striking veniremembers who- demonstrated that they would have difficulty "understanding] evidence" and, in particular, "assessing witness credibility.” 414 F.3d 288, 293, 300-01 (2d Cir.2005). She was especially concerned about striking veniremembers who might be reluctant to rely on one witness’ testimony to convict in drug cases (like Green’s), "because drug prosecutions often depend on the testimony of one witness.” Id. at 300-01. The voir dire transcript reflected that when the veniremember at issue was asked how she would judge witness credibility, she replied only "evidence”; furthermore, her "description of how she went about determining whether her son was lying to her” called into serious question her ability to serve in her role as a juror. Id. at 293. The court held that the district court had not erred in accepting this reason at both Steps Two and Three. Id. at 301. This prosecutor had a regular practice — she testified not that she preferred more thoughtful jurors, but that she struck veniremembers who demonstrated that they would have difficulty assessing witness credibility. Furthermore, she linked this (patently legitimate) practice to particular statements the veniremember made during voir dire.

26 . Vague preferences are particularly likely to conceal implicit bias, as the district judge — to his credit — -recognized. Prosecutors might well conceive of "life experience” in ways that have a profoundly disparate impact on members of different racial groups. Young black people may be less likely to enroll in college than young white people, but this can hardly be taken to signify that the average young black person has less "life experience” than the average young white person. Moreover, a vague preference may be more likely to play a part in a prosecutor’s decision to strike a veniremember who is black than it would if that juror were white, if the prosecu tor is motivated to a substantial degree by racial bias.

It is true (and most unfortunate) that Bat-son is not designed to root out implicit bias, as Justices Breyer and Marshall, along with one of our colleagues in the Northern District of Iowa, have discussed in some depth. See Miller-El v. Dretke (Miller-El II), 545 U.S. 231, 267-68, 125 S.Ct. 2317, 162 L.Ed.2d 196 (2005) (Breyer, J., concurring) (citing Batson, 476 U.S. at 106, 106 S.Ct. 1712 (Marhsall, J., concurring)); Mark W. Bennett, "Unraveling the Gordian Knot of Implicit Bias in Jury Selection: The Problems of Judge-Dominated Voir Dire, the Failed Promise of Batson, and Proposed Solutions,” 4 Harv. L. & Pol’y Rev. 149 (2010). However, the risk of implicit bias is acutely relevant when considering circumstantial evidence of the sort at issue here, because a prosecutor's jury selection approach offers no support at all for the state’s case if it is not consistently employed in a race-neutral fashion.

27 . In the course of this discussion, Van Stra-len mentioned, among other things, that R.O. "had not gone off to college.” If Van Stralen had instead testified that he employed a categorical practice of striking veniremembers who had not attended college, this would have been suspect. In Hernandez v. New York, the Supreme Court held:

"[A]n invidious discriminatory purpose may often be inferred from the totality of the relevant facts, including the fact, if it is true, that the [classification] bears more heavily on one race than another.” If a prosecutor articulates a basis for a peremptory challenge that results in the disproportionate exclusion of members of a certain race, the trial judge may consider that fact as evidence that the prosecutor's stated reason constitutes a pretext for racial discrimination. 500 U.S. 352, 363, 111 S.Ct. 1859, 114 L.Ed.2d 395 (1991) (quoting Washington v. Davis, 426 U.S. 229, 242, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976)). If an inference of discrimination can be drawn when a prosecutor testifies that he struck a'particular veniremember on a basis which, if applied universally, would result in disproportionate strikes of minority venire-members, that same inference is inescapable when a prosecutor informs the court that he in fact does have a consistent practice of striking veniremembers on that basis.

28 . Van Stralen described Juror Number 3's education and job experience as "positive” because they showed "initiative,” "intelligence,” "decision-making,” and that he "had something going for him.” But he did not assert that, R.O. lacked those qualities. Similarly, while Van Stralen suggested Juror Number 3 might have a favorable view of law enforcement, he did not contend that R.O. did not also have a positive view.