UNITED STATES of America, Plaintiff-Appellee, v. Daryle Lamont SELLERS, Defendant-Appellant

Concurrence by Judge Nguyen ;

Dissent by Judge Graber

NGUYEN, Circuit Judge:

Daryle Lamont Sellers was convicted of conspiracy to distribute cocaine and conspiracy to interfere with commerce by robbery after he was caught in a law enforcement reverse sting operation to rob a fictitious stash house. Sellers argues that he was targeted based on his race, and presents evidence that an overwhelming majority of the defendants targeted by law enforcement in similar investigations are African-Americans or Hispanics. To succeed on his selective enforcement claim, Sellers must show that the enforcement had a discriminatory effect and was motivated by a discriminatory purpose. He is unlikely to meet this demanding standard without information that only the government has. Sellers can obtain this information through discovery if he makes a threshold showing. We must decide what that showing is. We hold that in these stash house reverse-sting cases, claims of selective enforcement are governed by a less rigorous standard than that applied to claims of selective prosecution under United States v. Armstrong , 517 U.S. 456, 116 S.Ct. 1480, 134 L.Ed.2d 687 (1996).

BACKGROUND

In 2012, the Bureau of Alcohol, Tobacco, Firearms and Explosives (ATF) and Agent John Carr set up what is known as a stash house reverse-sting operation near downtown Los Angeles. These operations tend to follow a common format: An undercover agent poses as a disgruntled drug courier who is looking for help robbing the house where his employer is stashing (and guarding) a large quantity of drugs. The agent describes the stash house to individuals who have been targeted for the operation. Usually, the targets of stash house reverse-sting operations are identified using confidential informants. Informants are supposed to identify targets that have committed stash house robberies before or are capable of doing so.

The agent conducts a series of meetings with the targets and presents them with the opportunity to rob the stash house, and they devise a plan to do so. There is no stash house to rob, and there are no drugs-this is a reverse-sting, after all. But at the last meet-up, just before they are set to leave and carry out the plan, the targets are arrested for conspiracy to commit the robbery and associated crimes.

The details of the specific stash house reverse-sting operation here, for the most part, are irrelevant to Sellerss selective enforcement claim, and so we state them only in brief. In March 2012, a confidential informant staying at a hotel in a predominantly black and Hispanic area of Los Angeles targeted one of Sellerss co-defendants for a stash house reverse-sting operation, ostensibly because the informant believed that the co-defendant was involved in selling drugs. The co-defendant, who is black, was put in touch with Agent Carr, and the stash house reverse-sting was underway. On July 9, 2012, Sellers attended a planning meeting for the robbery with the co-defendant, Agent Carr, and others. Eventually, the stash house robbery was set for July 16, and, after one final meeting confirming the plan, the robbery crew (all of whom are black) was arrested.

Sellers and his co-defendants were indicted for conspiracy to possess and distribute cocaine, conspiracy to commit robbery, and possession of a firearm in furtherance of these crimes. Sellers moved to dismiss the indictment for outrageous government misconduct and sought discovery on a claim of selective enforcement. Sellers presented data collected by an attorney in the Central District of California showing that of 51 defendants indicted in stash house reverse-sting operations between 2007 and 2013, at least 39 were black or Hispanic. Similarly, Agent Carr testified that more than 55 of the approximately 60 individuals who have been indicted in his stash house reverse-sting operations are people of color. Relying on the standard set forth in Armstrong for obtaining discovery on selective prosecution claims, the district court denied the motion.

Sellers was convicted by a jury and sentenced to 96 months imprisonment. He timely appeals.

STANDARD OF REVIEW

We have jurisdiction pursuant to 28 U.S.C. § 1291. Whether the district court applied the correct discovery standard is a legal question that we review de novo. See United States v. Washington , 797 F.2d 1461, 1470 n.12 (9th Cir. 1986). We review the district courts determination that Sellers did not make the requisite discovery showing for abuse of discretion.

United States v. Arenas-Ortiz , 339 F.3d 1066, 1069 (9th Cir. 2003). The court necessarily abuses its discretion when it applies the wrong legal standard. See United States v. Hinkson , 585 F.3d 1247, 1261 (9th Cir. 2009) (en banc) (citing Cooter & Gell v. Hartmarx Corp. , 496 U.S. 384, 405, 110 S.Ct. 2447, 110 L.Ed.2d 359 (1990) ).

DISCUSSION

I.

We are not working from an entirely blank slate. Selective prosecution and selective enforcement claims are undoubtedly related, see Lacey v. Maricopa County , 693 F.3d 896, 920 (9th Cir. 2012) (en banc), and the Supreme Court addressed the threshold discovery showing required for selective prosecution claims over two decades ago in Armstrong . 517 U.S. at 458, 116 S.Ct. 1480. The question we face is whether Armstrong s standard is equally applicable to claims for selective enforcement, particularly in the stash house reverse-sting context. We first address Armstrong s discovery standard for selective prosecution cases and then explain why we join the Third and Seventh Circuits in declining to adopt it wholesale here.

A. Armstrong

To establish a claim of selective prosecution, a defendant must show both discriminatory effect and discriminatory purpose. Armstrong , 517 U.S. at 465, 116 S.Ct. 1480. In Armstrong , the Supreme Court consider[ed] the showing necessary for a defendant to be entitled to discovery on a claim of selective prosecution. Id. at 458, 116 S.Ct. 1480. The Court adopted a rigorous standard, id. at 468, 116 S.Ct. 1480, whereby a defendant must show that the Government has failed to prosecute others who are similarly situated to the defendant as evidence of discriminatory effect. Id. at 469, 116 S.Ct. 1480.

The Court explained its rationale for such a high standard. Id. at 464-68, 116 S.Ct. 1480. It observed that [i]n order to dispel the presumption that a prosecutor has not violated equal protection, a criminal defendant must present clear evidence to the contrary. Id. at 465, 116 S.Ct. 1480 (internal quotation marks omitted). The Court instructed us to be hesitant and not unnecessarily impair the prosecutors constitutional function. Id. (internal quotation marks omitted). And it was this justification[ ] for a rigorous standard for the elements of a selective-prosecution claim that require[d] a correspondingly rigorous standard for discovery in aid of such a claim. Id. at 468, 116 S.Ct. 1480.

Armstrong was thus premised on the notion that the standard for discovery for a selective prosecution claim should be nearly as rigorous as that for proving the claim itself. In other words, the standard was intentionally hewn closely to the claims merits requirements. See id. ; see also United States v. Hare , 820 F.3d 93, 99 (4th Cir. 2016) (The standard for obtaining discovery in support of a selective prosecution claim is only slightly lower than for proving the claim itself. (internal quotation marks omitted) ).

B. Material Differences Between Selective Prosecution and Selective Enforcement

Selective prosecution is not selective enforcement-especially not in the stash house reverse-sting context. There are two main differences that warrant departure from the Armstrong standard: First, law enforcement officers do not enjoy the same strong presumption that they are constitutionally enforcing the laws that prosecutors do. Second, the nature of reverse-sting operations means that no evidence of similarly situated individuals who were not targeted exists.

1. Presumption of Regularity

[T]he presumption of regularity supports ... prosecutorial decisions .... Armstrong , 517 U.S. at 464, 116 S.Ct. 1480 (internal quotation marks omitted). This presumption gives a measure of protection (and confidentiality) to prosecutors deliberative processes, which are covered by strong privileges. United States v. Davis , 793 F.3d 712, 720 (7th Cir. 2015) (en banc). Prosecutors occupy a special province of the executive branch and have broad discretion to enforce our nations laws, Armstrong , 517 U.S. at 464, 116 S.Ct. 1480 (internal quotation marks omitted).

On the other hand, [a]gents of the ATF and FBI are not protected by a powerful privilege or covered by a presumption of constitutional behavior. Davis , 793 F.3d at 720. Criminal defendants are allowed discovery for various aspects of law enforcement operations, including statements made and actions taken by investigating agents. Agents investigatory decisions are regularly questioned at trial, and their credibility is put before courts and juries. Thus, agents occupy a different space and role in our system than prosecutors; they are not charged with the same constitutional functions, and their decisions are more often scrutinized by-and in-courts.

Because the same presumption of regularity and deference to prosecutorial decision-making policy concerns do not apply in the selective enforcement context, we need not apply as rigorous a standard here.

2. Nonexistent Evidence

In the selective prosecution context, statistical evidence of differential treatment is ostensibly available. See Armstrong , 517 U.S. at 466-67, 470, 116 S.Ct. 1480. For instance, comparing who was arrested with who was prosecuted, or the demographics of those prosecuted in state and federal courts for the same crime, may evince differential treatment of similarly situated individuals. See id. That is not the case in the context of selective enforcement. Asking a defendant claiming selective enforcement to prove who could have been targeted by an informant, but was not , or who the ATF could have investigated, but did not , is asking him to prove a negative; there is simply no statistical record for a defendant to point to. Cf. Chavez v. Ill. State Police , 251 F.3d 612, 640 (7th Cir. 2001) (In a meritorious selective prosecution claim, a criminal defendant would be able to name others arrested for the same offense who were not prosecuted by the arresting law enforcement agency; conversely, plaintiffs who allege that they were stopped due to racial profiling would not, barring some type of test operation, be able to provide the names of other similarly situated motorists who were not stopped.).

This is especially true for stash house reverse-sting operations, where no independent crime is committed; the existence of the crime is entirely dependent on law enforcement approaching potential targets, and any comparative statistics can only be derived by the government and its informants choosing to approach and investigate white individuals. See Hare , 820 F.3d at 101 (In the stash house sting context, a defendant would face considerable difficulty obtaining credible evidence of similarly situated individuals who were not investigated by ATF.).

In Armstrong , the Supreme Court concluded that requiring evidence about similarly situated defendants would not make a selective-prosecution claim impossible to prove. That is not the case here; comparative statistics do not exist. As did the Court in Armstrong , we set the discovery standard accordingly and find that a lower standard is warranted under these circumstances.

C. Davis and Washington

The Third and Seventh Circuits have already come to the conclusion that Armstrong s rigorous discovery standard does not apply in the context of selective enforcement claims involving stash house reverse-sting operations. See United States v. Washington , 869 F.3d 193, 219-21 (3d Cir. 2017), cert. denied , --- U.S. ----, 138 S.Ct. 713, 199 L.Ed.2d 582 (2018) ; Davis , 793 F.3d at 719-21. The Fourth Circuit has described the arguments for doing so as well taken. Hare , 820 F.3d at 101 (citing Davis ). We are now the fourth circuit to address this question in the stash house reverse-sting context.

In United States v. Davis , the Seventh Circuit, sitting en banc, emphasized that Armstrong was about prosecutorial discretion and how federal prosecutors deserve a strong presumption of honest and constitutional behavior, which cannot be overcome simply by a racial disproportion in the outcome because disparate impact differs from discriminatory intent. 793 F.3d at 720. The court found that the sort of considerations that led to the outcome in Armstrong do not apply to a contention that agents of the FBI or ATF engaged in racial discrimination when selecting targets for sting operations, or when deciding which suspects to refer for prosecution. Id. at 721. Thus, based on the Davis defendants showing that 88 of the 94 defendants prosecuted after stash house reverse-sting operations in the district were black or Hispanic, the court held that information from supervisors or case agents of the FBI and ATF would be outside the scope of Armstrong and discoverable. Id. at 715, 721-22 ; see also id. at 722 (The racial disproportion in stash-house prosecutions remains troubling ... and it is a legitimate reason for discovery.).

In Washington , the Third Circuit discussed Davis at length and ultimately agree[d] with the Davis court that district judges have more flexibility, outside of the Armstrong [ ] framework, to permit and manage discovery on claims for selective enforcement related to stash house reverse-sting operations. 869 F.3d at 213. The court found that Armstrong was grounded in part on the special solicitude courts have shown to prosecutors discretion that does not inevitably flow to the actions of law enforcement. Id. at 216, 219. The court also took note of the defendants argument that the fact that there are likely to be no records of similarly situated individuals who were not arrested or investigated ... would transform the functional impossibility of Armstrong [ ] into a complete impossibility. Id. at 216. The court held that so long as the defendants proffer contains reliable statistical evidence, or its equivalent, ... a defendant need not, at the initial stage, provide some evidence of discriminatory intent, or show that ... similarly situated persons of a different race or equal protection classification were not arrested or investigated by law enforcement. Id. at 221. The court remanded for the district court to determine in the first instance whether the defendant, who had shown that all of the defendants prosecuted in connection with stash house reverse-sting operations in the district were black, was entitled to any additional discovery. Id. at 200, 222.

D. The Resulting Standard

Today we join the Third and Seventh Circuits and hold that Armstrong s rigorous discovery standard for selective prosecution cases does not apply strictly to discovery requests in selective enforcement claims like Sellerss. Contrary to Armstrong s requirements for selective prosecution claims, a defendant need not proffer evidence that similarly-situated individuals of a different race were not investigated or arrested to receive discovery on his selective enforcement claim in a stash house reverse-sting operation case. While a defendant must have something more than mere speculation to be entitled to discovery, what that something looks like will vary from case to case. The district court should use its discretion-as it does for all discovery matters-to allow limited or broad discovery based on the reliability and strength of the defendants showing.

II.

Having set forth the applicable standard, we turn to Sellerss threshold showing in this case. Sellers argues that the evidence he presented regarding the demographics of those indicted based on stash house reverse-sting operations entitles him to discovery on his selective enforcement claim. Because the district court applied an incorrect legal standard, we follow our normal practice of remanding to the district court to determine in the first instance whether Sellers has met the standard we outline today. See Kirkpatrick v. Chappell , 872 F.3d 1047, 1058 (9th Cir. 2017) (When a district court applies the wrong legal standard ..., we ordinarily remand the case so that it may apply the correct one in the first instance.). It may be that Sellers does not meet even a lower standard. Or it may be that he meets the standard but is entitled to no more discovery than he already received in connection with his entrapment defense. Or Sellers may be entitled to ask the government to be more forthcoming about its practices with regard to stash house reverse-sting operations. We leave that to the district court to decide.

The dissent, arguing that Sellers isnt entitled to discovery under any standard, purports to apply some lesser level of proof for a claim of selective enforcement, Dissent at 28, but then applies exactly the standard articulated in Armstrong for a claim of selective prosecution. The cases upon which it relies all involve selective prosecution claims. See United States v. Bass , 536 U.S. 862, 122 S.Ct. 2389, 153 L.Ed.2d 769 (2002) (per curiam) (involving claim of selective prosecution in seeking the death penalty); Arenas-Ortiz , 339 F.3d at 1068 (involving claim that the United States Attorney engaged in a pattern of selective prosecution of Hispanic males for illegal reentry); United States v. Turner , 104 F.3d 1180, 1181 (9th Cir. 1997) (involving contention that the defendants had been selected for prosecution on crack cocaine charges on racial grounds).

In conflating the standards for discovery in selective prosecution and selective enforcement claims, the dissent overlooks the main reason for distinguishing them: the presumption that prosecutors properly discharged their official duties absent clear evidence to the contrary. Armstrong , 517 U.S. at 464, 116 S.Ct. 1480 (quoting United States v. Chem. Found., Inc. , 272 U.S. 1, 14-15, 47 S.Ct. 1, 71 L.Ed. 131 (1926) ). Because [t]he justifications for a rigorous standard for the elements of a selective prosecution claim are not present in a selective enforcement claim, the latter does not require a correspondingly rigorous standard for discovery in aid of such a claim. Id. at 468, 116 S.Ct. 1480. Thus, obtaining discovery on a selective enforcement claim does not require some evidence tending to show the existence of [both] essential elements of the defense, discriminatory effect and discriminatory intent, id. at 468, 116 S.Ct. 1480 (quoting United States v. Berrios , 501 F.2d 1207, 1211 (2d Cir. 1974) ), notwithstanding that the defendant will eventually need to show both elements to prevail on the claim, see Lacey v. Maricopa County , 693 F.3d 896, 920 (9th Cir. 2012) (en banc). See Dissent at 28 ([A] litigant need not prove entitlement to relief in order to obtain discovery.). Therefore, even if the dissent were correct that Sellers presented no evidence of discriminatory effect, see Dissent at 31, evidence of discriminatory intent may be enough to warrant discovery.

CONCLUSION

The order denying discovery is VACATED and the case is REMANDED for limited post-judgment proceedings consistent with this opinion.

We described in detail one example of a stash house reverse-sting operation in United States v. Black , 733 F.3d 294, 298-301 (9th Cir. 2013).

These meetings are supposed to serve as a vetting process to ensure that the targeted individuals are willing and capable of committing the stash house robbery.

See 18 U.S.C. §§ 924(c)(1)(A), 1951 ; 21 U.S.C. § 846.

We address Sellerss appeal of the denial of his motion to dismiss for outrageous government conduct and challenges to his sentence in a simultaneously-filed memorandum disposition.

At times, Sellers has styled his claim as one of selective prosecution, but it is more properly considered a claim for selective enforcement since Sellers takes issue with how he was targeted at the outset of the operation. The district court considered it as such, and Sellers adopted this characterization on appeal. We follow suit.

No white defendants were identified; the remaining 12 were of unknown races.

We decline the governments invitation to apply a plain error standard of review. Sellers has consistently argued that there are meaningful differences between Armstrong and his case, and the district court expressly considered whether there are differences between selective prosecution and selective enforcement claims. Sellerss claim was properly presented in the district court, and we are free to address it. Lebron v. Natl R.R. Passenger Corp. , 513 U.S. 374, 379, 115 S.Ct. 961, 130 L.Ed.2d 902 (1995).

United States v. Arena-Ortiz , 339 F.3d 1066 (9th Cir. 2003), which involved a selective prosecution claim, does not foreclose our consideration of the difficulty of obtaining certain types of evidence. If the discovery standard for these types of claims had already been set, difficulty meeting the standard would not be a valid excuse for failing to do so. See id. at 1070-71. But it is not, and we have leeway when deciding the appropriate standard at the outset.

In Hare , the Fourth Circuit assume[d] that the defendants showing that all 32 of the defendants prosecuted in stash house reverse-sting cases in the district were black was sufficient to warrant discovery into selective enforcement but found that the defendants had already received all of the discovery to which they would be entitled. 820 F.3d at 98, 101.

Other tools in a district courts tool box (such as in camera review) may also aid the courts decision as to whether discovery is warranted.

Indeed, even in the selective prosecution context, the Supreme Court left open the possibility that direct admissions by prosecutors of discriminatory purpose (rather than the usual circumstantial evidence) would entitle the defendant to discovery without showing some evidence of discriminatory effect. Armstrong , 517 U.S. at 469 n.3, 116 S.Ct. 1480.

Sellerss conviction and sentence are otherwise unaffected by this remand. Sellerss conditional motion for remand (docket entry no. 61) is DENIED as moot.

NGUYEN, Circuit Judge, concurring:

For more than two decades, the government has engaged in the controversial practice of stash house reverse stings, in which the government feels compelled to invent fake crimes and imprison people for long periods of time for agreeing to participate in them, United States v. Black , 750 F.3d 1053, 1057 (9th Cir. 2014) (Reinhardt, J., dissenting from the denial of rehearing en banc). Despite widespread criticism of this tawdry and disreputable tactic, United States v. Lewis , 641 F.3d 773, 777 (7th Cir. 2011) (We use the word tawdry because the tired sting operation seems to be directed at unsophisticated, and perhaps desperate, defendants who easily snap at the bait put out for them by [the government agent].); United States v. Kindle , 698 F.3d 401, 414 (7th Cir. 2012) (Posner, J., dissenting in part), vacated on rehg en banc sub nom. United States v. Mayfield , 771 F.3d 417 (7th Cir. 2014), the government has expanded fake stash house sting operations from a single metropolitan area to cities nationwide.

While these operations do not ... reduc[e] the actual flow of drugs, the government touts them as an important tool to catch people inclined to commit home invasions. United States v. Hudson , 3 F.Supp.3d 772, 786 (C.D. Cal. 2014), revd sub nom. United States v. Dunlap , 593 F. Appx 619 (9th Cir. 2014). But when the government fails to target known criminal enterprises or people suspected of engaging in serious crimes, the practice is highly questionable and raises troubling questions about race-based targeting.

There is no legitimate dispute that these stings primarily affect people of color, but the government has steadfastly resisted any defense attempt to determine whether enforcement is racially biased. Courts exercising their gatekeeping role in determining whether discovery is warranted should recognize that the choice of locations for these operations may have evidentiary significance to a claim of discriminatory effect and discriminatory intent.

I.

Stash house reverse stings have been widely criticized on a number of race-neutral grounds. See United States v. Conley , 875 F.3d 391, 402 (7th Cir. 2017) (commenting on the substantial body of criticism of similar stash house cases both from this circuit and others); United States v. Washington , 869 F.3d 193, 197 (3d Cir. 2017) ([R]everse sting operations have grown increasingly controversial over the years, even as they have grown safer and more refined.); see also United States v. Flowers , 712 F. Appx 492, 509 (6th Cir. 2017) (Stranch, J., concurring) (This concerning ... tactic has rightly drawn criticism in news reporting, scholarly writing, and from the judiciary.). See generally Marc D. Esterow, Note, Lead Us Not into Temptation: Stash House Stings and the Outrageous Government Conduct Defense , 8 Drexel L. Rev. Online 1, 28-33 (2016).

To begin with, the government need not pursue existing criminal enterprises or individuals suspected of involvement in any crime-let alone stash house robberies. Indeed, the government typically outsources the selection of a target to a confidential informant, introducing a host of biases and bad incentives into the process. See, e.g. , United States v. Black , 733 F.3d 294, 303 (9th Cir. 2013) (observing that the government was trolling for targets when the confidential informant provocatively cast his bait in places defined only by economic and social conditions); see also United States v. McLean , 199 F.Supp.3d 926, 943 (E.D. Pa. 2016) (citing the inherently arbitrary way in which stash house sting cases first ensnare suspects as a reason that enforcing a ... mandatory minimum would offend due process); United States v. Cambrelen , 29 F.Supp.2d 120, 125-26 (E.D.N.Y. 1998) (finding the use of confidential informants especially troubling since those people are often in the process of negotiating down their own drug sentences or charges with prosecutors, and have enormous incentive to inflate the drug quantities involved in the cases they help prosecute), affd , 5 F. Appx 30 (2d Cir. 2001).

Not surprisingly, given the way in which they are selected, targets of stash house stings and their co-conspirators sometimes have modest criminal résumés. See, e.g. , United States v. McKenzie , 656 F.3d 688, 692 (7th Cir. 2011) (The crime proposed was ... a massive one; it is somewhat baffling, then, that the young men who the authorities recruited did not have massive criminal histories to match.); see also United States v. Brown , 299 F.Supp.3d 976, 987 (N.D. Ill. 2018) (ATF does not always target existing criminal enterprises or have prior suspicion of potential targets, and instead the stings often ensnare low-level crooks who jump at the bait of a criminal windfall. (quoting Flowers , 712 F. Appx at 509 (Stranch, J., concurring) ) ). In such cases, the government is creating hardened criminals out of individuals who might otherwise lead productive lives.

The danger of ... reverse stings is substantially heightened when the government takes aim at poor neighborhoods and tempts their residents with the prospect of making large amounts of money through criminal activity. At the right moment and when described in attractive enough terms, such offers may lead astray otherwise law abiding young men living in poverty, and motivate them to make false or exaggerated claims about their qualifications to serve as participants in the proposed venture-including claims about prior criminal experience that lack any substantial basis in truth.

Black , 750 F.3d at 1056 (Reinhardt, J., dissenting from the denial of rehearing en banc) (internal citation omitted).

Another serious problem with fictional stash house operations is that the government has virtually unfettered ability to inflate the amount of drugs supposedly in the house and thereby obtain a greater sentence for the defendant. United States v. Briggs , 623 F.3d 724, 729 (9th Cir. 2010). It is no coincidence that in reverse stings across the country, the amount of the hypothetical cocaine to be stolen is always purported to exist in quantities exceeding five kilograms, the amount that triggers a mandatory 10-year minimum sentence. Esterow, supra , at 29. In addition, targets are often encouraged to bring items, such as guns, zip ties, or duct tape, that will not only serve as evidence of their intent to participate in the conspiracy, but will also allow the charging of additional crimes. Eda Katharine Tinto, Undercover Policing, Overstated Culpability , 34 Cardozo L. Rev. 1401, 1447-48 (2013).

Controlling the fictitious amount of drugs allows the government to enhance not only the targets sentencing exposure but also the attractiveness of joining the conspiracy in the first place. See Hudson , 3 F.Supp.3d at 786 ([T]he Government must make the robbery scheme tempting enough to nab a potential criminal. The Government thus sets the drug amount at a level ... that no poverty-ridden individual could pass up. ... [T]his ruse is not meant to simply skim off those individuals likely to commit similar crimes; rather, it is designed to never fail.). Similarly, the government can minimize the obstacles that a defendant must overcome to obtain the drugs, Briggs , 623 F.3d at 730, such as by making the stash house guards insignificant in number or potency. The ease with which the government can manipulate these factors makes us wary of such operations in general, id. , yet we continue to approve of them, no matter how egregious.

II.

The governments stated rationale for stash house stings is to protect normal neighborhoods from the armed crime associated with the drug trade. A normal neighborhood, as the agent in this case explained, is a middle-class neighborhood without security bars on the doors and windows-in other words, safe and relatively affluent. Stash houses are often placed in such neighborhoods to avoid drawing the suspicion of law enforcement, and the risk of a stash house robbery endangers any innocent family living nearby.

Keeping neighborhoods safe from violent crime is laudable, but the benefits and burdens of stash house stings fall along racial lines. For reasons that transcend law enforcement, the comfortable neighborhoods being protected are overwhelmingly white. See, e.g. , Steven Raphael & Melissa Sills, Urban Crime, Race, and the Criminal Justice System in the United States , in A Companion to Urban Economics 515, 516 (Richard J. Arnott & Daniel P. McMillen eds., 2006) ([W]ithin large metropolitan areas, the residents of poor, largely minority neighborhoods suffer [from crime] disproportionately.).

More troublingly, law enforcement agents-whether consciously or not-appear to primarily target racial minorities. Nationwide, approximately 90% of the individuals currently imprisoned as a result of [a] ... stash house sting are either African-American or Hispanic. Esterow, supra , at 31. This consequence naturally flows from operations conducted almost exclusively in minority neighborhoods. Here, for example, the agent acknowledged that he conducted stings in an area of Los Angeles that he considered to be predominantly African American and Hispanic rather than whiter and wealthier neighborhoods. As a result, in the agents cases that led to prosecutions, fewer than five of the roughly sixty defendants were white. [A]ctions having foreseeable and anticipated disparate impact are relevant evidence to prove the ultimate fact, forbidden purpose. Columbus Bd. of Ed. v. Penick , 443 U.S. 449, 464, 99 S.Ct. 2941, 61 L.Ed.2d 666 (1979).

In examining what constitutes evidence of discriminatory effect, there is a significant difference between selective enforcement and selective prosecution. To show that similarly situated individuals of other races were not prosecuted, a defendant would need to present evidence that individuals of other races were potentially liable for prosecution and that prosecutors knew this but did not act on it-a difficult but not impossible task. See Armstrong , 517 U.S. at 470, 116 S.Ct. 1480 (For instance, respondents could have investigated whether similarly situated persons of other races were prosecuted by the State of California and were known to federal law enforcement officers, but were not prosecuted in federal court.). Because prosecutors do not themselves investigate crimes, they are limited to prosecuting only individuals whom law enforcement agents have identified as probable criminals. Prosecutors discretion, though substantial, is finite.

Law enforcement agents, on the other hand, do not deal with a closed universe of criminal suspects. When conducting a reverse sting, literally anyone could be a target. See Black , 733 F.3d at 315 (Noonan, J., dissenting) (In the population of this country, there is an indefinite number of persons who dream of clever and unlawful schemes to make money. Does their dreamy amorality cast them all as fit candidates for a sting by their government?).

There is no reason to suspect that persons of a particular race are more likely to agree to commit a stash house robbery unless one believes that persons of that race are inherently more prone to committing violent crime for profit-a dangerously racist view that has no place in the law. If law enforcement agents target potential stash house robbers in a race-neutral way, then the racial breakdown of targeted individuals would presumably closely mirror that in the community. If it doesnt, then thats potentially indicative that the agents or their informants are using discriminatory procedures.

In Armstrong , the Supreme Court expressed concern with the presumption that people of all races commit all types of crimes without considering the premise that any type of crime is the exclusive province of any particular racial or ethnic group. 517 U.S. at 469, 116 S.Ct. 1480 (quoting United States v. Armstrong , 48 F.3d 1508, 1516-17 (9th Cir. 1995) ). To support its assertion that some crimes are committed primarily by individuals of a particular race, the Court cited presumably reliable statistics showing the racial composition of convicted perpetrators of three crimes. Id. at 469-70, 116 S.Ct. 1480.

In the selective enforcement context, extrapolating the incidence by race of particular crimes (or, as here, the propensity to commit particular crimes) from conviction rates makes sense only if police investigate crime in a racially unbiased manner. But all too often that isnt true. See, e.g. , Emma Pierson et al., A large-scale analysis of racial disparities in police stops across the United States (2017), https://5harad.com/papers/traffic-stops.pdf (finding that black drivers are stopped more often than white drivers relative to their share of the driving-age population, that blacks and Hispanics are more likely to be ticketed, searched, and arrested than similarly situated white drivers, and that blacks and Hispanics are searched on the basis of less evidence than whites).

We have found that facially neutral policies ha[ving] a foreseeably disproportionate impact on an identifiable group do not amount to an equal protection violation. Lee v. City of Los Angeles , 250 F.3d 668, 687 (9th Cir. 2001), But I question whether conducting stash house operations almost exclusively in neighborhoods known to be black and Hispanic, and excluding neighborhoods known to be white, is in fact a facially neutral policy. See Washington v. Davis , 426 U.S. 229, 241, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976) (A prima facie case of discriminatory purpose may be proved ... by the absence of [minorities] on a particular jury combined with the failure of the jury commissioners to be informed of eligible [minority] jurors in a community, or with racially non-neutral selection procedures. (internal citations omitted) ); cf. McCleskey v. Kemp , 481 U.S. 279, 294, 107 S.Ct. 1756, 95 L.Ed.2d 262 (1987) (rejecting statistical evidence of racial disparity in death penalty sentences to show discriminatory intent because each particular decision to impose the death penalty is made by a petit jury selected from a properly constituted venire). Even if, for the sake of argument, stash house robberies are more likely to be committed by persons of color than by whites for reasons having nothing to do with race, limiting reverse stings to minority neighborhoods will still result in the systematic overrepresentation of minority targets.

For example, consider a region with a population that is two-thirds white and one-third minority and in which 0.002% of minorities and 0.001% of whites commit stash house robberies. Although minorities in this fictitious region are twice as likely as whites to commit stash house robberies, there are twice as many whites as minorities in the population as a whole. Consequently, the region has equal numbers of minorities and whites who are stash house robbers. If law enforcement agents use race-neutral procedures to identify potential stash house robbers, then half of the targets should be minorities and half should be whites. But if agents limit their sting operations to neighborhoods where minorities comprise 80% of the population, then eight minorities will be targeted for every white target.

III.

Evidence that law enforcement agents or their confidential informants scoured disproportionately minority neighborhoods in search of stash house reverse sting targets is evidence of discriminatory effect. If the agents knew they were limiting their operations to minority neighborhoods and made no effort to stage operations elsewhere, without more, thats also potentially indicative of discriminatory purpose. Whether this is enough evidence in this case to entitle Sellers to additional discovery is for the district court to resolve in the first instance.

Like many of my colleagues across the country, I am greatly disturbed by the governments practice and, in particular, its disproportionate impact on people of color. The governments use of stash house reverse stings warrants closer scrutiny.

I concur in the memorandum disposition regarding the remaining claims.

One can question the correctness of the Supreme Courts holdings on this point, but it is hard to quarrel with their clarity. And, as a circuit court, we are bound to follow a controlling Supreme Court precedent until it is explicitly overruled by that Court, whether we agree with its reasoning or not. Nunez-Reyes v. Holder , 646 F.3d 684, 692 (9th Cir. 2011) (en banc).

In addition, the Courts extrapolation assumed that the judicial system is unbiased. Yet it is well documented that defendants of color, African Americans in particular, are more likely to be convicted than similarly situated white defendants. See, e.g. , Shamena Anwar et al., The Impact of Jury Race in Criminal Trials , 127 Q.J. Econ. 1017 (2012); Sheri Lynn Johnson, Black Innocence and the White Jury , 83 Mich. L. Rev. 1611 (1985).

To be clear, I wholly reject the notion that persons of color are inherently more likely to commit certain crimes, i.e. , that race or ethnicity is a causal factor. There may be causal factors that are correlated with race, leading to a higher incidence of perpetrators among certain races. For example, if poverty is a causal factor of stash house robberies and wealth is distributed unequally by race for unrelated reasons-such as a history of racial oppression-then, setting aside other causal factors, persons of races with a less-than-equal share of the communitys wealth will commit stash house robberies at a greater rate than persons of other races. Yet if the wealth inequality were remedied, then the racial disparity among stash house robbers would disappear.

GRABER, Circuit Judge, dissenting:

I respectfully dissent. In my view, we need not-and therefore should not-opine about the standard for obtaining discovery in selective enforcement cases because, under either a high or a low standard, Defendants evidentiary proffer is wanting as a matter of law.

I

The district court denied Defendants motion for discovery on the claim of selective enforcement. We review that decision for abuse of discretion. United States v. Arenas-Ortiz , 339 F.3d 1066, 1069 (9th Cir. 2003). An error of law is, of course, one form of abuse of discretion. Koon v. United States , 518 U.S. 81, 100, 116 S.Ct. 2035, 135 L.Ed.2d 392 (1996). But here, the district court relied on correct legal principles and did not otherwise commit an abuse of discretion.

I would assume, without deciding, that the high bar for obtaining discovery for a claim of selective prosecution , enunciated in United States v. Armstrong , 517 U.S. 456, 116 S.Ct. 1480, 134 L.Ed.2d 687 (1996), does not apply to a motion to obtain discovery for a claim of selective enforcement . That is, I would assume that some lesser level of proof is required in order to obtain discovery for a claim of selective enforcement, as the majority opinion now holds. The basic elements of the two types of claims-discriminatory effect and discriminatory intent-are the same, as we held in Lacey v. Maricopa County , 693 F.3d 896, 920 (9th Cir. 2012) (en banc); but a litigant need not prove entitlement to relief in order to obtain discovery. Rather, the question in each case is whether there is enough evidence before the court to suggest that further discovery is warranted. In both the prosecution and the enforcement contexts, as well as under general discovery principles, the court is not required to grant discovery with respect to a speculative claim. Rivera v. NIBCO, Inc. , 364 F.3d 1057, 1072 (9th Cir. 2004).

The district court denied the motion primarily on the ground that there was insufficient evidence of discriminatory effect to raise an inference justifying discovery, rejecting Defendants statistical proffer. The district courts reasoning and result are correct. The Supreme Court has instructed-an instruction that we have followed-that raw statistics alone are not at all probative of discriminatory effect. That is, they are irrelevant .

In United States v. Bass , 536 U.S. 862, 122 S.Ct. 2389, 153 L.Ed.2d 769 (2002) (per curiam), the defendant argued that the government had elected to seek the death penalty, rather than a lesser punishment, because of his race. In support of his request for discovery, the defendant presented nationwide statistics showing that the government charged African-Americans with death-eligible offenses more than twice as often as it charged whites with death-eligible offenses. Id. at 863, 122 S.Ct. 2389. The Supreme Court held that those statistics did not entitle the defendant to discovery, not because the numbers were insufficiently probative to warrant further exploration but-more categorically-because raw statistics regarding overall charges say nothing about charges brought against similarly situated defendants . Id. at 864, 122 S.Ct. 2389 (first emphasis added). In other words, the raw statistics were not just unpersuasive; they were irrelevant .

We have treated similar statistical evidence the same way in cases both before and after Bass -cases that the district court cited in reaching the conclusion that Defendants evidence fell short. In United States v. Turner , 104 F.3d 1180, 1184-85 (9th Cir. 1997), for example, we held that statistics pertaining to the number of overall convictions for crack cocaine charges brought against different groups did not entitle the defendants to discovery. We relied on Armstrong , explaining that sheer numbers, without further evidence, d[o] not advance a defense of selective prosecution. Id. A few years later, in Arenas-Ortiz , 339 F.3d at 1070, we explained similarly that raw statistics are probative only if paired with other evidence.

Those precedents, and Armstrong itself, do not teach simply that generalized statistical evidence is frowned on in this context. Nor do they teach that raw statistics are insufficient only in cases involving prosecutors. Rather, they stand for the principle that raw statistics concerning the racial makeup of a group of defendants (or, here, a group of suspects targeted by a law enforcement agency), without other evidence, are irrelevant to proving the existence of a discriminatory effect.

Given those precedents, I conclude that the district court correctly rejected Defendants evidence as insufficient to create an inference of discriminatory effect, because that evidence consisted only of non-comparative raw statistics. The evidence is insufficient whether we apply either a rigorous Armstrong standard or the more forgiving standard devised by the majority opinion. Because the statistics that Defendant presented say nothing, Bass , 536 U.S. at 864, 122 S.Ct. 2389, they cannot, under any standard, entitle Defendant to demand discovery.

Moreover, it is possible for a defendant to find some comparative statistical information that would satisfy the Supreme Courts requirements. In this case, for instance, the district court ordered production of the ATF manual insofar as it describes how to determine which persons to pursue as potential targets. Suppose that the manual states that agents should pursue people who have been convicted of distributing large quantities of heroin, or people who have been released from prison in the past six months after having been incarcerated for robbery. In theory, at least, a defendant could ascertain who is in the universe of intended targets and determine whether there is disproportionate enforcement.

But even if practical considerations made it impossible to find such information, our precedent would not allow discovery. We considered and rejected just such a challenge in Arenas-Ortiz , 339 F.3d at 1070-71. There, the defendant argued that the district court had erred in denying his request for discovery because it would have been an insuperable task to produce the requisite evidence. Id. As we explained in rejecting that argument, it is in the nature of a standard that there will be times when that standard cannot be met. Merely demonstrating that better evidence cannot be obtained without discovery does not suddenly render otherwise insufficient evidence sufficient. Id. at 1071 ; see also Armstrong , 517 U.S. at 470, 116 S.Ct. 1480 (rejecting this courts concern about evidentiary obstacles that defendants would face if required to produce evidence of differential treatment of similarly situated members of other races).

In summary, we need not decide whether the standard for obtaining discovery on a selective enforcement claim is more generous than Armstrong s standard for a selective prosecution claim because the generalized evidence that Defendant offered is insufficient to raise any inference of discriminatory effect. The district court therefore permissibly denied discovery for that reason. Accordingly, the majority opinion is, at best, a gratuitous exercise and, at worst, an advisory opinion.

II

The majority opinion misconstrues Armstrong by conflating its two separate holdings. Nothing in the Courts discussion of the level of proof appropriate to a particular claim undercuts its holding that non-comparative statistical evidence is no proof at all.

In Armstrong , the Supreme Court reversed our circuits allowance of discovery in aid of a selective prosecution claim. To understand the Supreme Courts decision in context, it is useful to begin with what we held and, thus, what the Supreme Court rejected.

Our court held en banc that a defendant who seeks discovery in connection with a claim of selective prosecution need only demonstrate a colorable basis for believing that wrongful discrimination took place. United States v. Armstrong , 48 F.3d 1508, 1510 (9th Cir. 1995) (en banc). The district court granted the defendants motion for discovery when the defendants presented a study showing that all 24 cases closed in 1991 and handled by the Federal Public Defenders Office for the Central District of California, in which a particular drug crime was charged, involved African-American defendants. Id. at 1511. The district court ruled that the statistical data raised a question about the motivation of the Government sufficient to justify discovery to reveal the prosecutors criteria for bringing that and similar cases. Id. at 1512. We agreed, id. at 1515-19, and emphasized that

statistical disparities alone may suffice to provide the evidence of discriminatory effect and intent that will establish a prima facie case of selective prosecution. ... [W]e hold that inadequately explained evidence of a significant statistical disparity in the race of those prosecuted suffices to show the colorable basis of discriminatory intent and effect that warrants discovery on a selective prosecution claim.

Id. at 1513-14 (citations and footnote omitted). We expressly rejected a requirement for defendants to compile facts which are not easily obtainable by them, such as the racial breakdown and offense characteristics of defendants represented by other counsel. Id. at 1514. The concurring opinion stated that, at the discovery stage, only some evidence, tending to show selective prosecution, is required. Where there is evidence of a large enough number of prosecutions directed at a single race over a sufficiently long period of time, eventually there becomes a point where that evidence is sufficient to establish a colorable basis of selective prosecution. Id. at 1521 (Wallace, J., concurring). With respect to the role of prosecutors, we reasoned that the broad discretion they possess over charging decisions means that they may be the only source of information demonstrating discrimination, thereby justifying a generous standard for discovery. Id. at 1514 (majority).

In summary, we held: (1) Only a colorable basis for concluding that unlawful discrimination occurred is required to support discovery. One rationale for that standard (repeated by the majority opinion here, pp. 10-11) is that it may be hard for defendants to obtain information, which is largely in prosecutors hands. (2) Raw statistics, without comparative numbers, sufficed to demonstrate a colorable basis.

Against that backdrop, I read the Supreme Courts decision to follow the same two-part structure. The Court first discussed the presumption that prosecutors discharge their official duties properly, but also reaffirmed that a prosecutors decision to pursue a case may not, under the Equal Protection Clause, be based on race. Armstrong , 517 U.S. at 463-66, 116 S.Ct. 1480. Rather than supporting this courts loose discovery standard, though, the role of the prosecutor justifies a high standard: the showing necessary to obtain discovery should itself be a significant barrier to the litigation of insubstantial claims. Id. at 464, 116 S.Ct. 1480. The standard should be rigorous and can be described with a variety of phrases, including substantial threshold showing. Id. at 468, 116 S.Ct. 1480.

After establishing a standard, the Court went on to consider, second, whether the statistical information supplied by the defendants sufficed; the Court answered that separate question in the negative. The Court summarized its evidentiary holding as follows:

In this case we consider what evidence constitutes some evidence tending to show the existence of the discriminatory effect element. The Court of Appeals held that a defendant may establish a colorable basis for discriminatory effect without evidence that the Government has failed to prosecute others who are similarly situated to the defendant. We think it was mistaken in this view. The vast majority of the Courts of Appeals require the defendant to produce some evidence that similarly situated defendants of other races could have been prosecuted, but were not, and this requirement is consistent with our equal protection case law. As the three-judge panel explained, [s]elective prosecution implies that a selection has taken place.

The Court of Appeals reached its decision in part because it started with the presumption that people of all races commit all types of crimes-not with the premise that any type of crime is the exclusive province of any particular racial or ethnic group. It cited no authority for this proposition, which seems contradicted by the most recent statistics of the United States Sentencing Commission. ... Presumptions at war with presumably reliable statistics have no proper place in the analysis of this issue.

Id . at 469-70, 116 S.Ct. 1480 (citations omitted); see also id. at 465, 116 S.Ct. 1480 (To establish a discriminatory effect in a race case, the claimant must show that similarly situated individuals of a different race were not prosecuted. (emphasis added) ); id. at 467, 116 S.Ct. 1480 ([The defendants] urge that cases such as Batson v. Kentucky , 476 U.S. 79, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986), and Hunter v. Underwood , 471 U.S. 222, 105 S.Ct. 1916, 85 L.Ed.2d 222 (1985), cut against any absolute requirement that there be a showing of failure to prosecute similarly situated individuals. We disagree.).

Turning to the statistics that the defendants offered to the district court, the Court concluded that their study did not constitute some evidence tending to show the existence of the essential elements of a selective-prosecution claim. The study failed to identify individuals who were not black and could have been prosecuted for the offenses for which [the defendants] were charged, but were not so prosecuted. Id . at 470, 116 S.Ct. 1480 (citations omitted).

In short, the Supreme Court in Armstrong held: (1) To justify discovery concerning a claim of selective prosecution, a defendant must present some evidence tending to show a discriminatory effect and a discriminatory purpose. One rationale for that rigorous standard is the presumption that prosecutors perform their duties properly. (2) Raw statistics, without comparative numbers, are inadequate as a matter of law to permit an inference of discriminatory effect.

Thus, Armstrong contains two distinct holdings. The first sets a standard for how persuasive a litigants showing must be to justify discovery. The second rejects, categorically, the relevance of raw statistics. Ratcheting the standard up or down has no bearing on the separate, second holding, which rests not on the nature of prosecutors , as the majority opinion avers, pp. 10, 15-16, but on the nature of evidence that the Court deemed irrelevant. And it is entirely clear from the Courts discussion that the second, evidentiary point (including the requirement to produce comparative information) applies to all equal protection claims. Armstrong , 517 U.S. at 465-68, 116 S.Ct. 1480.

III

Finally, the majority writes that, even if my dissenting opinion is correct in concluding that Defendant presented no evidence of discriminatory effect, evidence of discriminatory intent may be enough to warrant discovery. Maj. op. at 856. Whatever other function that suggestion serves, it cannot justify discovery here because there is no evidence of discriminatory intent, either.

As the district court accurately observed, Defendant offered no independent evidence of discriminatory intent. Rather, Defendant asserted that the ATFs adoption of policies that had a discriminatory effect necessarily demonstrated a racial animus. Given the absence of legally sufficient evidence of discriminatory effect, this bootstrapping attempt fails.

Indeed, the record contradicts a claim of subjective discriminatory purpose on the governments part with respect to Defendant. ATF agents testified that they pursued violent offenders with experience in the drug trade. More importantly, the governments enforcement effort did not target Defendant at all. Instead, a co-conspirator recruited him into the conspiracy.

For the foregoing reasons, I respectfully dissent.

The Bureau of Alcohol, Tobacco, and Firearms (ATF) devised this scheme in Miami in the early 1990s. Drug cartels moving huge quantities of cocaine through South Florida attracted freelance criminals who tried to poach the shipments, often resulting in shootouts or attacks on innocent people. Brad Heath, ATF Uses Fake Drugs, Big Bucks to Snare Suspects , USA Today, June 26, 2013, https://www.usatoday.com/story/news/nation/2013/06/27/atf-stash-houses-sting-usa-today-investigation/2457109. Over the next two decades, reverse stash house stings proliferated, with operations in at least 22 states. Id.

A reverse sting both eliminates one potential stash house robber ... and deters other criminals from joining stash house robberies .... The greater security that fictitious stash house stings confer on real stash houses ... reduces their cost of self-protection, which is a principal cost of the illegal-drug business. Kindle , 698 F.3d at 416 (Posner, J., dissenting in part).