CHRISTEN, Circuit Judge:

Sontos Maudilia Diaz-Reynoso, a native and citizen of Guatemala, petitions for review of the Board of Immigration Appeals’ (BIA) order dismissing her appeal of an Immigration Judge’s (IJ) order denying her application for withholding of removal and relief under the Convention Against Torture (CAT). Diaz-Reynoso seeks withholding of removal based on her fear that she would be persecuted in Guatemala on account of her membership in the particular social group of “indigenous women in Guatemala who are unable to leave their relationship.” Diaz-Reynoso argues she is entitled to relief under CAT because, if returned to Guatemala, the Guatemalan government would acquiesce in torture she would suffer at the hands of her husband.

On her withholding claim, the BIA concluded that Diaz- Reynoso’s proposed particular social group was not cognizable, relying on the Attorney General’s decision in Matter of A-B-, 27 I. & N. Dec. 316 (A.G. 2018). With respect to Diaz-Reynoso’s CAT claim, the BIA concluded that Diaz-Reynoso failed to establish that the government of Guatemala would acquiesce in any torture she might suffer. Because the BIA’s decision on both bases for relief departs from its own precedent and is contrary to this court’s case law, we grant the petition for review and remand for further proceedings.

I

Diaz-Reynoso was born in 1989 in the small, rural town of Yamoj, in the Guatemalan highlands. She is a member of DIAZ-REYNOSO V. BARR 7

the indigenous group known as Mam. In 2008, Diaz-Reynoso moved in with a man named Arnoldo Vasquez-Juarez, who is also Mam. Although they did not legally marry, Diaz- Reynoso and Vasquez-Juarez had a common-law marriage and Diaz-Reynoso refers to Vasquez-Juarez as her husband.

Vasquez-Juarez subjected Diaz-Reynoso to physical and sexual abuse. Among other things, he forced Diaz-Reynoso to work in the coffee fields without pay, and to have sex with him. When Diaz-Reynoso did not comply with his demands, Vasquez-Juarez attacked her, hitting her on her head and all over her body, sometimes with a belt. Diaz-Reynoso testified that she was attacked weekly, and that the resulting bruises sometimes lasted for eight to ten days.

In 2012, after four years of living with Vasquez-Juarez, Diaz-Reynoso fled and entered the United States without documentation. She was apprehended, and after roughly a month in detention, returned to Guatemala. Diaz-Reynoso moved back in with her family in Yamoj.

As soon as Diaz-Reynoso returned, Vasquez-Juarez came to find her. Vasquez-Juarez told Diaz-Reynoso that if she did not return to live with him, he would kill her, kill her daughter,1 or harm her mother. Diaz-Reynoso returned to live with Vasquez-Juarez for about a year. The abuse got worse during that time. Diaz-Reynoso then escaped and went to live with a friend in another town for roughly a year. She was in hiding during this period and did not leave her friend’s house. After that, Diaz-Reynoso returned to her family home In articulating why a remand is required, the majority first explains that “[t]he BIA seems to have erroneously understood Matter of A-B- to forbid any mention of feared harm within a proposed social group.” Maj. Op. 19–20. The majority consequently “disagree[s]” that “in order to exist independently from the petitioner’s feared harm, a proposed group may not refer to that harm at all.” Id. at 23. This reflects legal error and departs from the decisions of our sister circuits. See, e.g., Gonzales-Veliz, 938 F.3d at 232 (holding that a proposed group of “Honduran women unable to leave their relationship” was not cognizable because it was “defined by, and d[id] not exist independently of, the harm—i.e., the inability to leave”).

Simply stated, a proposed group that incorporates harm within its definition is not a group that “exist[s] independently of the harm asserted in an application for asylum or statutory withholding of removal.” Matter of A-B-, 27 I. & N. Dec. at 334 (emphasis added) (quotations omitted). Recall that the overall framework requires an asylum or withholding applicant to prove three things: (1) the existence of a cognizable particular social group; (2) membership in that group; and (3) “a risk of persecution on account of” (or “because of”) “his membership.” Reyes, 842 F.3d at 1132 n.3 (emphasis added).

DIAZ-REYNOSO V. BARR 51

Now imagine a proposed social group that is transparently based in part on the harm, such as “Guatemalan men who are harmed by gangs.” Saying that a petitioner was harmed by gangs based on his membership in the group “Guatemalan men who are harmed by gangs” would collapse the inquiries. And it would create the very circularity problem that Matter of A-B- sought to avoid in the domestic violence context, allowing nexus to be proven by the group definition itself. See Matter of A-B-, 27 I. & N. Dec. at 334 (citing, e.g., Matter of M-E-V-G-, 26 I. & N. Dec. at 236 n.11, 243; Matter of W-G-R-, 26 I. & N. Dec. at 215). The majority thus errs in claiming that I “do not explain why a person seeking relief on the basis of membership in a particular social group should be required to omit any mention of threatened persecution.” Maj. Op. 28. The reason, as I have explained, is the anti- circularity principle.

Contrary to the suggestion in the majority opinion, Matter of A-B- did not endorse the concept of a particular social group that references harm in its definition. Maj. Op. 23–25. Matter of A-B- in fact confirms that such a group is impermissible. Matter of A-B- expressly rejected the proposed social group in Matter of A-R-C-G-, which was “married women in Guatemala who are unable to leave their relationship.” Matter of A-B-, 27 I. & N. Dec. at 319. This proposed group obviously does contain some immutable characteristics not defined by the harm, namely, “married women in Guatemala.” But the Attorney General concluded the proposed social group still incorporated a reference to harm, because a claimed “inability to leave” was “effectively defined to consist of women in Guatemala who are victims of domestic abuse.” Id. at 335. For that reason, the group violated the anti-circularity principle and was “not cognizable.” Id. at 334.

52 DIAZ-REYNOSO V. BARR

If referring to harm within the group definition were allowed, Matter of A-B- on this central point should have come out the other way. It is thus true, as the majority explains, that Matter of A-B- did not “identify a categorical ‘circularity problem.’” Maj. Op. 20 (emphasis added). But that is only because the anti-circularity rule had long existed in the law. Matter of A-B- thus did not “identify” and announce this rule, but it certainly applied it. Whether the anti-circularity rule applies is determined on a case-by-case basis by examining the proposed social group that an applicant brings forward. See Matter of A-B-, 27 I. & N. Dec. at 340; Matter of M-E-V-G-, 26 I. & N. Dec. at 251. But when the rule does apply, it is, indeed, a categorical one.

Matter of A-B-’s reliance on Rreshpja v. Gonzales, 420 F.3d 551 (6th Cir. 2005), only confirms this point. The Attorney General in Matter of A-B- specifically cited Rreshpja for the rule that “[t]he individuals in the group must share a narrowing characteristic other than their risk of being persecuted.” 27 I. & N. Dec. at 335 (alteration in original) (quoting Rreshpja, 420 F.3d at 556). In Rreshpja, the court understood the proposed social group to be “young (or those who appear to be young), attractive Albanian women who are forced into prostitution.” 420 F.3d at 555. The Sixth Circuit rejected this group because “a social group may not be circularly defined by the fact that it suffers persecution.” Id. at 556 (emphasis added).

The proposed particular social group in Rreshpja was not completely defined by the harm, but the court rejected it because it was still partially defined by the harm. Id. The majority therefore gets Rreshpja backwards. Maj. Op. 23. The Attorney General in Matter of A-B- was not purporting to allow a group defined in part by the harm by relying on a

DIAZ-REYNOSO V. BARR 53

Sixth Circuit opinion that rejected a proposed social group that was itself partially defined by the harm, and that therefore suffered from the same problem as the proposed group in this case.

Notably, both petitioner and her amicus agree with my interpretation of Matter of A-B-. Petitioner herself regards the “holding” of Matter of A-B- to be that a particular social group “cannot be defined, even in part, by the fact of persecution.” (Emphasis added). And amicus Center for Gender & Refugee Studies likewise agrees that under Matter of A-B-, the circularity rule “require[s] a group to be defined completely independently of the harm.” (Emphasis in original). Petitioner and the amicus both regard this as a new rule, but it is not. It is the same rule that has always applied: “[t]o be cognizable, a particular social group must exist independently of the harm asserted.” Matter of A-B-, 27 I. & N. Dec. at 334 (emphasis in original) (quotations omitted); see also Matter of M-E-V-G-, 26 I. & N. Dec. at 236 n.11.

Unlike the majority, my interpretation of Matter of A-B- is also aligned with that of other circuits. The majority opinion “recognize[s] that, consistent with Matter of A-B-, numerous courts have deemed proposed social groups that referred to feared harm to be impermissibly circular.” Maj. Op. 31. The concession is well-taken. But it is, if anything, a considerable understatement, because other circuits have routinely denied petitions for review presenting materially identical proposed social groups to the one at issue here, that also referenced harm in the group definition itself.

In Gonzales-Veliz v. Barr, 938 F.3d 219 (5th Cir. 2019), the Fifth Circuit denied a petition for review seeking asylum

54 DIAZ-REYNOSO V. BARR

and withholding of removal based on the proposed social group “Honduran women unable to leave their relationship.” Id. at 223. The Fifth Circuit explained that “under A-B-’s analysis, [this] group cannot constitute a particular social group” because “[t]he group is defined by, and does not exist independently of, the harm—i.e., the inability to leave.” Id. at 232. In other words, the Fifth Circuit held, this group was not cognizable because it was “impermissibly defined in a circular manner.” Id.

Several months later, the Eleventh Circuit issued a similar decision in Amezcua-Preciado v. U.S. Attorney General, 943 F.3d 1337 (11th Cir. 2019) (per curiam). There, the Eleventh Circuit denied a petition for review seeking relief based on the proposed social group “women in Mexico who are unable to leave their domestic relationships.” Id. at 1339. The Eleventh Circuit recognized that the petitioner’s “proposed social group suffers from the kinds of problems the Attorney General identified in A-B- as likely to render most groups of victims of private violence not cognizable.” Id. at 1344. And the Eleventh Circuit made clear that “to the extent [the petitioner’s] proposed group of Mexican women . . . are unable to leave their domestic relationships because they fear physical or psychological abuse by their spouse or domestic partner, this group is defined by the underlying harm asserted.” Id. at 1345. In reading Matter of A-B- to allow proposed social groups that include feared harm in the group definition, the majority departs from both the reasoning and the results of the Fifth and Eleventh Circuit decisions.

Indeed, the majority’s decision is inconsistent with many cases from other circuits that have denied petitions for review that advanced proposed social groups materially identical to the one advanced here. See, e.g., Perez-Agustin v. U.S.

DIAZ-REYNOSO V. BARR 55

Attorney Gen., 798 F. App’x 608, 609 (11th Cir. 2020) (per curiam) (rejecting as impermissibly circular proposed social group of “indigenous women from Guatemala, who are native Mam speakers, who are victim[s] of sexual violence”) (alteration in original); Martinez Casco v. U.S. Attorney Gen., 800 F. App’x 835, 838 (11th Cir. 2020) (per curiam) (rejecting as impermissibly circular proposed social group of “female domestic violence victims who are unable to leave”); Serrano-de Portillo v. Barr, 792 F. App’x 341, 342–43 (5th Cir. 2020) (per curiam) (rejecting as impermissibly circular proposed social group of “El Salvadoran women targeted by gang members to be gang girlfriends”); Garcia-Ventura v. Barr, 788 F. App’x 969, 970–71 (5th Cir. 2019) (per curiam) (rejecting as impermissibly circular proposed social groups of “victims of domestic violence at the hands of their domestic partner and unable to leave their domestic partner” and “victims of domestic violence who are viewed as property by virtue of their positions within a domestic relationship”); Gonzalez-De Moreira v. U.S. Attorney Gen., 787 F. App’x 659, 662 (11th Cir. 2019) (per curiam) (rejecting as impermissibly circular proposed social groups of “Salvadoran women who are victims of violence” and “Salvadoran children who are victims of violence”); Reyes v. Sessions, 750 F. App’x 656, 659 (10th Cir. 2018) (rejecting as impermissibly circular proposed social group of “female victims of domestic violence”). Indeed, our court in an unpublished decision rejected as impermissibly circular a proposed group of “women who have been harassed and threatened by men and whose complaints to police have failed to result in protection,” because the group was “defined by the claimed persecutory conduct” and thus did “not exist

56 DIAZ-REYNOSO V. BARR independently from the claimed persecution.” Calderon- Espejo v. Barr, 771 F. App’x 371, 372–73 (9th Cir. 2019).2

The majority acknowledges this “plethora of unpublished decisions that rejected groups similar to the one advanced here,” but claims it “cannot discern the reasoning of those decisions simply from their outcome.” Maj. Op. 32 & n.9. Setting aside that the Fifth and Eleventh Circuits have rejected proposed social groups materially identical to the one here in published opinions, see Amezcua-Preciado, 943 F.3d at 1345; Gonzales-Veliz, 938 F.3d at 232, the numerous unpublished decisions reaching the same result easily reveal their straightforward reasoning: the proposed group in each case failed because it included a reference to harm, thus presenting the circularity issue.

The groups in these cases were defined in part by immutable characteristics that were separate and apart from

2 Even before Matter of A-B-, courts also routinely rejected proposed social groups as circular if they referred to alleged harm or persecution. See e.g., Soto-Ambrocio v. Sessions, 724 F. App’x 456, 458 (6th Cir. 2018) (per curiam) (rejecting as impermissibly circular proposed social group of “young women from Guatemala subject to abuse from families”); Moreno v. Lynch, 628 F. App’x 862, 865 (4th Cir. 2015) (rejecting as impermissibly circular proposed social group of “Hondurans who have been targeted by the police and their criminal associates to engage in drug trafficking”); Rreshpja, 420 F.3d at 555–56 (rejecting as impermissibly circular proposed social group of “young (or those who appear to be young), attractive Albanian women who are forced into prostitution”). The majority is thus not correct that the government and this dissent (and by implication, the other circuits) “confuse the definition of a ‘particular social group’ with one of its components—i.e., the group’s shared immutable characteristic.” Maj. Op. 27. The requirement that a proposed social group exist independent of the harm is its own separate requirement. Matter of M-E-V-G-, 26 I. & N. Dec. at 236 n.11.

DIAZ-REYNOSO V. BARR 57

the harm, such as women from a certain country. But they were also defined in part by an “inability to leave” or some similar formulation that was regarded as a reference to harm. Unlike today’s opinion, however, none of these cases treated the BIA decisions on review as having “erroneously understood Matter of A-B- to forbid any mention of feared harm within a proposed social group.” Maj. Op. 19–20. Instead, they treated the “mention” of that harm as a dispositive violation of the longstanding anti-circularity principle.^[3] The majority gives two primary reasons for departing from the case law of other circuits and allowing proposed

3 The majority is correct that the D.C. Circuit recently stated (in a case challenging a policy memorandum not at issue here) that the notion that “the group must be ‘separate’ from the harm, not consisting of the harm, even in part,” is “flatly inconsistent” with Matter of A-B-. Grace v. Barr, — F.3d —, 2020 WL 4032652, at[*16] (D.C. Cir. 2020). However, the D.C. Circuit premised this point on an asserted government concession as to whether the phrase “inability to leave” could be a reference to something other than harm, id., which is a different issue (the majority addresses this issue in its Part V.C and I address it in Part II.C below). In Grace, the D.C. Circuit held that the policy memorandum at issue there correctly stated the “circularity rule” as set forth in Matter of A-B-. Id. The court described the “circularity rule” as requiring that “social groups must ‘exist independently’ of the harm claimed by the applicant, that is, the applicant must be able to establish the group’s existence ‘without defining [it] by the fact of persecution.’” Id. at[*14] (quoting Matter of A-B-, 27 I. & N. Dec. at 334) (alteration in original). To the extent Grace “endorsed the view that a proposed social group is not disqualified if it includes mention of feared persecution” as the majority claims, Maj. Op. 29, the D.C. Circuit offered no explanation for this point, which is both incorrect and against the weight of authority applying Matter of A-B-, as detailed in this dissent.

58 DIAZ-REYNOSO V. BARR social groups that include a reference to harm. Both rationales fail as a matter of law.

First, the majority holds that a proposed social group definition can include a reference to harm because “persecution may be relevant to a group’s social distinction.” Maj. Op. 20; see also id. at 23–24. This reflects a misunderstanding of BIA precedent and the conflation of two independent requirements for a “particular social group” to be cognizable: (1) that the group exist independent of the harm and (2) that it be socially distinct in the relevant society. These are different requirements. Indeed, in explaining the “social distinction” element, the BIA in Matter of M-E-V-G- pointed out that while “the social group must exist independently of the fact of persecution,” this is a separate criterion from “social distinction.” Matter of M-E-V-G-, 26 I. & N. Dec. at 236 n.11.

The majority dismisses this as “mechanistic[],” Maj. Op. 27, but these separate requirements perform meaningful analytical work. It is certainly true that persecution can be used to demonstrate that a proposed group is “socially distinct.” As the BIA has explained, “[t]he act of persecution by the government may be the catalyst that causes the society to distinguish [a proposed social group] in a meaningful way and consider [it] a distinct group.” Matter of M-E-V-G-, 26 I. & N. Dec. at 243. But this does not mean that the definition of the group itself can include the harm. Why? Because the entire reason the group “must exist independently of the harm asserted,” Matter of A-B-, 27 I. & N. Dec. at 334 (emphasis in original) (quotations omitted), is to avoid the circularity problem that would otherwise allow petitioners automatically to fulfill the separate nexus requirement by injecting the harm

DIAZ-REYNOSO V. BARR 59 itself into the very definition of the group whose membership is supposedly causing the persecution.

The majority’s hypotheticals only prove this point. Maj. Op. 25–26. The first one, used in Matter of M-E-V-G-, consists of the proposed particular social group “former employees of a country’s attorney general.” Matter of M-E-V-G-, 26 I. & N. Dec. at 242. The second hypothetical is found in United Nations High Commissioner for Refugees (UNHCR) Guidelines and involves the proposed social group of left-handed men. UNHCR, Guidelines on International Protection, ¶ 14, U.N. Doc. HCR/GIP/02/02 (May 7, 2002) [hereinafter UNHCR Guidelines]. Critically, unlike the proposed social group in this case, neither hypothetical involves a proposed group defined in any way by the harm itself.

But though the fact of persecution does not define the group, these hypotheticals show that the harm may permissibly be used to show the group is socially distinct in the relevant society. Take “former employees of a country’s attorney general” as an example. As the BIA explained, “[a]lthough such a shared past experience is immutable and the group is sufficiently discrete, the employees may not consider themselves a separate group within the society.” Matter of M-E-V-G-, 26 I. & N. Dec. at 242. But if “the government begins persecuting them,” “[u]pon their mistreatment, it is possible that these people would experience a sense of ‘group,’ and society would discern that this group of individuals, who share a common immutable characteristic, is distinct in some significant way.” Id. at 242–43.

60 DIAZ-REYNOSO V. BARR

That is consistent with the anti-circularity principle, because while “[t]he act of persecution by the government may be the catalyst that causes the society to distinguish the former employees in a meaningful way and consider them a distinct group,” “the immutable characteristic of their shared past experience exists independent of the persecution.” Id. at 243 (emphasis added). The UNHCR made the same point about its “[l]eft-handed men” hypothetical: the group would still be defined by “the attribute of being left-handed and not the persecutory acts.” UNHCR Guidelines ¶ 14 (emphasis added).

The majority’s two hypotheticals thus demonstrate the very analytical distinction the majority sweeps aside. Just because persecution may show that a group is socially distinct does not mean the group itself can be defined in some way by that persecution. The majority apparently suggests it would create no issue to define the hypothetical group of “former employees of the attorney general” as “former employees of the attorney general, who are being hunted down and killed.” Maj. Op. 26. But that is exactly the type of group courts have routinely rejected under the anti-circularity principle. See ante at 53–56 & n.2. The majority is thus incorrect in suggesting that I overlook that BIA precedent “unequivocally establishes that a group’s persecution may be relevant to” the social distinction requirement. Maj. Op. 27. Persecution is relevant to the requirement of social distinction, as I have explained. But that does not mean groups can include in their definitions a reference to harm, contrary to the separate

DIAZ-REYNOSO V. BARR 61

requirement that the group exist independent of the harm. See, e.g., Matter of M-E-V-G-, 26 I. & N. Dec. at 236 n.11.4

Second, and relying on snippets of language in BIA decisions, the majority holds that a particular social group can include harm in its definition because “[t]he BIA [has] explained that a ‘social group cannot be defined exclusively by the fact that it is targeted for persecution.’” Maj. Op. 21 (quoting Matter of C-A-, 23 I. & N. Dec. 951, 960 (BIA 2006) (emphasis in original)). The majority quotes other BIA decisions that use this “exclusively” language, but these decisions in turn rely on the BIA’s earlier opinion in Matter of C-A-. See id. at 21–22, 31.

The majority’s reliance on “exclusively” is not correct. That a proposed social group cannot be defined “exclusively” by the harm does not somehow mean it could then be based in part on the harm. If that were true, Matter of A-B-—which rejected a proposed social group that was not based exclusively on the harm—would have come out the other way on this critical issue. And if the only limit was that a group cannot be based “exclusively” on the harm, there would have

4 The majority’s hypothetical of Tutsis fleeing Rwanda is inapt. Maj. Op. 28–29. There, the proposed social group would be defined based on ethnicity, not harm. See Donchev v. Mukasey, 553 F.3d 1206, 1220 (9th Cir. 2009). A person who experienced genocide based on ethnicity would have no reason to define his proposed social group by including the harm in the group definition itself. Petitioner here, by contrast, had understandable strategic reasons to try that approach here, as I explain below. The majority is correct that the anti-circularity rule is not specific to the domestic violence context. But that makes the majority’s departure from the rule more problematic, not less. Under the majority’s approach, the many cases that have rejected proposed social groups that referenced harm were all wrongly decided on this point.

62 DIAZ-REYNOSO V. BARR

been no reason for the BIA repeatedly to articulate the principle that the group “must exist independently of the harm asserted.” Matter of A-B-, 27 I. & N. Dec. at 334 (emphasis altered) (quotations omitted); Matter of M-E-V-G-, 26 I. & N. Dec. at 236 n.11. Indeed, cases the majority cites that use the “exclusively” formulation also state the rule that the group must exist independent of the harm. See Maj. Op. 22 n.5 (citing Perez-Rabanales, 881 F.3d at 67; Paloka, 762 F.3d at 196).

I suppose it is possible to imagine a proposed social group that is defined “exclusively” by harm (e.g., “victims of physical assaults”). But most proposed social groups that have the circularity problem are defined in part by the harm and in part by something else, such as gender or nationality. There is nothing in the BIA’s precedent suggesting it did not intend those proposed social groups to come within the anti- circularity principle, whose logic clearly extends to them. To the contrary, the BIA’s formulation—“must exist independently”—ensures that groups defined in part by the harm are not allowed.

The origins of the “exclusively” phrasing also demonstrate it was not intended to operate in the way the majority uses it, i.e., implicitly to allow groups defined in part by the harm. As noted, this “exclusively” language can be traced to the BIA’s decision in Matter of C-A-. But there, the BIA stated that “the [UNHCR] Guidelines state that ‘a social group cannot be defined exclusively by the fact that it is targeted for persecution,’” although “persecutory action toward a group may be a relevant factor in determining the visibility of a group in a particular society.” Matter of C-A-, 23 I. & N. Dec. at 960 (emphasis in original) (quoting UNHCR Guidelines ¶¶ 2, 14).

DIAZ-REYNOSO V. BARR 63

The quoted language drives home the distinction that the majority elides, which is that persecution can be relevant for social distinction (previously known as social visibility), but not for the definition of the group itself. The quoted language also shows that the majority is misusing the “exclusively” phrasing on its own terms. The BIA in Matter of C-A- used this language when quoting the UNHCR Guidelines, which I discussed earlier. Matter of C-A-, 23 I. & N. Dec. at 960. And these Guidelines likewise make clear that a group cannot be defined in part by the harm.

In the Guidelines, the UNHCR used the left-handed men hypothetical to show that persecution can be relevant to what we now call social distinction. UNHCR Guidelines ¶ 14; see Matter of M-E-V-G-, 26 I. & N. Dec. at 228. But the UNHCR made clear that in that circumstance, “it would be the attribute of being left-handed and not the persecutory acts that would identify them as a particular social group.” UNHCR Guidelines ¶ 14. That is a rejection of the majority’s view that a group can be defined in part by the harm.

The majority opinion is thus incorrect in claiming that I “ignore the word ‘exclusively’—along with similar limiting language in the BIA’s decisions and those from our sister circuits.” Maj. Op. 24. Instead, as the extensive discussion above confirms, it is the majority that is incorrectly using the “exclusively” formulation in service of authorizing groups that refer to harm, contrary to other circuits that have addressed substantially identical proposed social groups.

B

Equally mistaken is the majority’s suggestion that the BIA erred in this case because it is possible to bracket out a

64 DIAZ-REYNOSO V. BARR

reference to harm in the definition of a proposed social group in considering whether petitioner has put forward an “otherwise cognizable” group. Maj. Op. 21–22, 27, 39–40; see also id. at 20 (“If a group is otherwise cognizable, Matter of A-B- does not demand that it be devoid of any reference to an applicant’s claimed persecution.”) (emphasis added). The majority’s analysis suggests that we may consider some elements of a proffered social group as relevant to the immutable characteristic inquiry, consider other elements for their relevance to social distinction, and do so without impermissibly rewriting the petitioner’s proposed social group. See Maj. Op. 27–28.

That is not correct. I am aware of no case, and the majority cites none, that would allow a court (or the BIA) to take a petitioner’s proposed social group, break it down into constituent elements, and then assign those elements out to the distinct requirements that a petitioner must meet to demonstrate a cognizable social group. In my respectful view, that is not how this area of law works.

BIA precedent is clear that the agency must evaluate the proposed social group exactly as the petitioner has defined it. In precedent the majority cites, Maj. Op. 27, 28, 39 n.12, the BIA has held that “[w]here an applicant raises membership in a particular social group as the enumerated ground that is the basis of her claim, she has the burden to clearly indicate ‘the exact delineation of any particular social group(s) to which she claims to belong.’” Matter of W-Y-C- & H-O-B-, 27 I. & N. Dec. 189, 191 (BIA 2018) (quoting Matter of A-T-, 25 I. & N. Dec. ^[4], 10 (BIA 2009)). The Immigration Judge (IJ) must then analyze “the specific group” identified, and “[i]f an applicant is not clear as to the exact delineation of the proposed social group, the Immigration Judge should seek

DIAZ-REYNOSO V. BARR 65

clarification.” Id. On appeal, the BIA may only consider the specific proposed social group that the petitioner presented to the IJ. Id. at 191–92.

These various requirements are not empty formalities. Instead, they follow from the fact that a petitioner bears the burden of proof in demonstrating “the existence of a cognizable particular social group” and persecution because of her membership in that group. Reyes, 842 F.3d at 1132 n.3 (quoting Matter of W-G-R-, 26 I. & N. Dec. at 223); see also Cantarero-Lagos v. Barr, 924 F.3d 145, 151 (5th Cir. 2019) (“Requiring asylum and withholding applicants to delineate their [particular social group] to an IJ is simply a logical extension of this burden of proof.”).

Requiring the petitioner to delineate her proposed social groups has important practical benefits as well. It ensures that appropriate fact-finding is conducted at the IJ level. Matter of W-Y-C- & H-O-B-, 27 I. & N. Dec. at 191; Cantarero-Lagos, 924 F.3d at 152. And it creates a “rational administrative process” that allows the government to respond to a petitioner’s proposed social group, while preventing repeated remands to the IJ. Cantarero-Lagos, 924 F.3d at 152.

Consistent with this scheme, we have repeatedly determined that we lack jurisdiction to consider a new or different proposed social group that the petitioner did not present to the agency. See, e.g., Lozano Fuerte v. Barr, 804 F. App’x 742, 742–43 (9th Cir. 2020); Nesta-Najar v. Barr, 804 F. App’x 604, 606 (9th Cir. 2020); Perez Perez v. Barr, 804 F. App’x 597, 598 (9th Cir. 2020); Vasquez-Leon v. Barr, 804 F. App’x 612, 613 (9th Cir. 2020); Romero- Castro v. Barr, 790 F. App’x 64, 65 (9th Cir. 2020); Lopez-

66 DIAZ-REYNOSO V. BARR

Velasquez v. Sessions, 742 F. App’x 195, 196 n.1 (9th Cir. 2018).

The majority agrees that “courts cannot rewrite proposed social groups.” Maj. Op. 27. But the majority’s suggestion that we or the BIA can splice out elements of a proposed group and consider them for some, but not other purposes, is inconsistent with established precedent and practice. The same can be said of petitioner’s analogous suggestion that we should “look[] beyond the language used to describe [her] proposed social group.” (Emphasis added). Petitioners can propose multiple and alternative particular social groups to the agency; many petitioners do so. E.g., Honcharov v. Barr, 924 F.3d 1293, 1295 (9th Cir. 2019) (per curiam). But under the authority set forth above, we cannot take a petitioner’s proposed group and alter it. Indeed, the very idea that there are severable elements in the proposed social group is misplaced. There are no such elements; there is only the proposed group.

By the same token, if we subtract one element (here “inability to leave”), we likewise create a different group than the one the petitioner articulated. Tellingly, in none of the cases cited above denying petitions with similar proposed groups to the one here did the courts proceed to consider a version of the proposed group minus the impermissibly circular reference to the harm. Nor did those courts remand to the BIA with instructions to do this. The majority opinion thus fails to recognize that the definition of the proposed group—as the petitioner has defined it—is central to how this entire scheme works.

The majority errs in suggesting it can refashion petitioner’s proposed group because “[t]he BIA has taken

DIAZ-REYNOSO V. BARR 67

pains to state that ‘the shared trait of persecution does not disqualify an otherwise valid social group.’” Maj. Op. 24 (quoting Matter of M-E-V-G-, 26 I. & N. Dec. at 243). The BIA in Matter of M-E-V-G- was not implying that if a group is defined to include the harm, that the BIA (or a court) could carve out the reference to harm to produce an “otherwise valid social group.” Instead, the BIA said that if the petitioner had come forward with an “otherwise valid social group,” using the fact of persecution to demonstrate the social distinction requirement would not invalidate the group. Matter of M-E-V-G-, 26 I. & N. Dec. at 243.

Equally mistaken is the majority’s assertion that the Attorney General in Matter of A-B- “faulted the BIA[]” in Matter of A-R-C-G- for not considering the petitioner’s proposed social group without its reference to harm. Maj. Op. 29. In Matter of A-B-, the Attorney General was certainly not instructing the BIA to ignore a reference to harm and to come up with an “otherwise cognizable” social group using some of the other characteristics the petitioner brought forth.

To the contrary, when the Attorney General in Matter of A-B- faulted the BIA in Matter of A-R-C-G- for “avoid[ing] considering whether [the petitioner] could establish the existence of a cognizable particular social group without defining the group by the fact of persecution,” it was because the government in Matter of A-R-C-G- “conceded” that the petitioner there was a member of a cognizable social group. Matter of A-B-, 27 I. & N. Dec. at 334. The BIA’s error in Matter of A-R-C-G- was in accepting that concession without any further inquiry. Id. And had the BIA “properly analyzed the issues,” Matter of A-B- went on, it would have been “clear that the particular social group was not cognizable.” Id.

68 DIAZ-REYNOSO V. BARR

(emphasis added). As a result, whether petitioner here included an “inability to leave” in her proposed group by mistake or not, it is part of the group as she has defined it, and such a group is not valid under the anti-circularity principle. Id. at 334–35.

Of course, it is highly doubtful that the petitioner’s inclusion of “unable to leave” in her group definition was a mistake. Had petitioner omitted this phrase, and advanced only the remaining portions of her proposed group (“indigenous women in Guatemala”), that may have made it more difficult for her to meet other requirements for a particular social group, most notably “particularity” and “social distinction.” As the First Circuit recently recognized, the inclusion of “‘unable to leave’ in the group definition” was likely a response to “[s]ome case law [that] gave rise to a fear that ‘women,’ or ‘women in country X,’ or even ‘women in a domestic relationship,’ might be too large or too indistinct a group to serve as a particular social group.” De Pena-Paniagua v. Barr, 957 F.3d 88, 95 (1st Cir. 2020) (collecting cases); see also Matter of A-B-, 27. I. & N. Dec. at 336 (explaining that applicants sought “to avoid particularity issues by defining a narrow class—such as ‘Guatemalan women who are unable to leave their domestic relationships where they have children in common’”).

In fact, it was Matter of A-R-C-G-—the very case that Matter of A-B- overruled—that “held out ‘unable to leave’ as a supposedly smaller, better-suited safe harbor for women seeking asylum and withholding of removal.” De Pena- Paniagua, 957 F.3d at 95. In other words, the petitioner here almost certainly defined her proposed social group as she did to match what the BIA in Matter of A-R-C-G- then regarded as cognizable. The problem is that the Attorney General has

DIAZ-REYNOSO V. BARR 69

since overruled Matter of A-R-C-G- and held “that the particular social group [there] was not cognizable.” Matter of A-B-, 27 I. & N. Dec. at 334.

In short, the majority’s suggestion that we can ignore the petitioner’s reference to harm in her proposed social group is contrary to settled law. The same is true of the majority’s determination that we can extract the reference to harm from the definition of the group and treat is as relevant only to the social distinction requirement. Today’s decision breaks new ground in suggesting that any of this is permissible.

C

The majority’s analysis in Parts V.A–B treated a reference to harm in the definition of a proposed social group as permissible under Matter of A-B-. In Part V.C, the majority concludes that the BIA erred in dismissing petitioner’s appeal because the BIA “assumed her inability to leave her relationship was attributable to domestic violence,” when “[t]here are many reasons a petitioner might be unable to leave a relationship, including a variety of cultural, societal, religious, economic, or other factors.” Maj. Op. 33–34, 34 n.11 (quotations omitted). The majority holds that the BIA “avoided the case-specific inquiry demanded by Matter of A-B-” and failed to “carefully analyze[]” petitioner’s claim because it “assumed that domestic violence was the only reason [petitioner] was unable to leave her relationship.” Id. at 33–34, 36.

I will explain why this is mistaken momentarily. But it is first important to point out that this rationale is inconsistent with the rationales in Parts V.A–B. The import of Part V.C is that if “inability to leave” were based on the harm, the BIA

70 DIAZ-REYNOSO V. BARR

could have denied petitioner’s appeal on this basis as impermissibly circular. By the logic of this rationale, the problem instead is that here “inability to leave” may not be based on harm but on other factors, and the BIA erred by not considering that possibility.

But if a particular social group can reference harm in the group definition, then a remand for the reasons given in Part V.C is unnecessary. Under Parts V.A–B of the majority opinion, the BIA should be instructed to treat the proposed group as written as non-circular and to analyze the group only for its compliance with the other requirements (e.g., particularity and social distinction). If the BIA can just ignore the reference to harm in the proposed group (as a mistake or otherwise) and treat it as relevant only to social distinction or some other requirement, there is no reason for the BIA to now consider whether “inability to leave” may be a reference to something other than the harm alleged. Instead, the BIA should be instructed to excise “inability to leave” and consider whether the proposed social group is “otherwise cognizable.”

Or we can think of it this way: imagine the BIA had not made the supposed error that the majority identifies in Part V.C of its opinion and had even more explicitly regarded “inability to leave” as based on harm. In that circumstance, wouldn’t the BIA still have erred under Parts V.A–B of the majority opinion? The answer apparently is “yes,” because the majority says that one of the “problems with the BIA’s reasoning” below is that “the BIA misunderstood Matter of A-B-’s holding.” Maj. Op. 33. That reasoning would seem to necessitate a remand separate and apart from any particular arguments petitioner may have advanced to the BIA about

DIAZ-REYNOSO V. BARR 71 whether her “inability to leave” was based on something other than domestic violence. Id. at 33.

Regardless, the majority is wrong to hold the BIA erred by “assum[ing]” that petitioner’s “inability to leave her relationship was attributable to domestic violence” and “avoid[ing] the case-specific inquiry demanded by Matter of A-B- and the BIA’s precedents.” Maj. Op. 33, 36. Petitioner herself recognizes in her opening brief that in Matter of A-B-, the Attorney General “specifically rejected the applicant’s proposed social group of ‘El Salvadoran women who are unable to leave their domestic relationships where they have children in common with their partners.’” (Emphasis added). Applying Matter of A-B-, which the court today upholds under Chevron, the BIA here simply considered petitioner’s specific articulation of her proposed social group under the governing standards.

The BIA recognized that the petitioner’s proposed social group was “substantially similar to the group in Matter of A-R-C-G-.” The BIA at this point cited and incorporated the decision of the IJ, which made clear that the entire basis for petitioner’s proposed social group was her fear of domestic violence. For this reason, the BIA could (and did) determine that as in Matter of A-B-, “inability to leave” here was “effectively defined to consist of women . . . who are victims of domestic abuse because the inability ‘to leave’ was created by harm or threatened harm.” Matter of A-B-, 27 I. & N. Dec. at 334; see also Larita-Martinez v. INS, 220 F.3d 1092, 1096 (9th Cir. 2000) (explaining “the presumption” that the BIA “did review the evidence”).

The BIA thus concluded that—because petitioner’s group definition contained the same “inability to leave”

72 DIAZ-REYNOSO V. BARR

reference—it “suffer[ed] from the same circularity problem articulated by the Attorney General in Matter of A-B-.” This was not a “shortcut[],” as the majority claims, Maj. Op. 33, but an application of Matter of A-B- of the type that has been routinely affirmed in other circuits. Nor did the BIA offer a “one-sentence analysis,” as the majority claims. Maj. Op. 32. The BIA opinion devoted several paragraphs to this issue and also incorporated two pages of the IJ’s decision.

The majority is thus incorrect in asserting that the BIA here “committed the very same error it made in Matter of A- B-.” Maj. Op. 35. Again, the fatal misstep identified in Matter of A-B- was that the BIA in Matter of A-R-C-G- relied on concessions to hold a similar proposed group was cognizable. Matter of A-B-, 27 I. & N. Dec. at 334. In contrast, the BIA here did the opposite of what the BIA did in Matter of A-R-C-G-, because here it evaluated petitioner’s proposed social group under the longstanding anti-circularity principle that Matter of A-B- reaffirmed. It is hard to understand why the BIA was required to say more when petitioner’s claim clearly failed for a specific reason. As the Fifth Circuit explained in denying a petition with a materially identical proposed social group to the one here, “[a]s an adjudicatory body, the BIA necessarily relies on established precedents to decide matters pending before it and to avoid re-inventing the wheel every time.” Gonzales-Veliz, 938 F.3d at 232. And the “BIA did not blindly apply A-B- as a categorical ban” when “A-B-’s substantive reasoning happened to squarely foreclose [the petitioner’s] group.” Id.

Of course, even allowing that the vague phrase “unable to leave” could be based on something other than harm (such as cultural factors), I am hard-pressed to understand how the BIA was supposed to derive that nuance from petitioner’s

DIAZ-REYNOSO V. BARR 73

brief to the BIA. Relying on Matter of A-R-C-G-, petitioner’s counseled brief in the BIA argued that the proposed social group “‘married women in Guatemala who are unable to leave their relationship’ may share an immutable trait, where specific facts demonstrated a woman’s inability to leave her abusive marriage.” (Emphasis added). Petitioner’s brief to the BIA then discussed the “repugnant abuse” and “weekly beatings” at issue in Matter of A-R-C-G- and compared them to petitioner’s own experience, explaining that petitioner “[s]imilarly” “demonstrated that she suffered deplorable harm and abuse by her ‘common law’ husband.” Contrary to the majority opinion, the BIA thus did not “pluck[] one fact identified by [petitioner]—that her husband physically abused her.” Maj. Op. 35. This abuse was the central thrust of petitioner’s entire submission to the BIA.

The majority opinion nevertheless states that the petitioner “advanced evidence of economic, societal, and cultural factors that also may have prevented her from leaving her relationship.” Maj. Op. 34. But what the majority cites for this are pages from the petitioner’s brief to the IJ, which extensively discuss how the “inability to leave” is based on domestic violence. To the extent petitioner and her expert cited societal and cultural factors in Guatemala before the IJ, they repeatedly intermixed them with domestic violence itself. In other words, petitioner’s own application thus specifically wove domestic violence into its identification of the proposed group, at a time when Matter of A-R-C-G- was the prevailing BIA precedent. But in all events, petitioner certainly did not present “inability to leave” as a concept distinct from the harm itself. And presenting the proposed group in that way would likely have created problems for other aspects of petitioner’s required showing, such as social distinction and nexus.

74 DIAZ-REYNOSO V. BARR

Under all these circumstances, I cannot fault the BIA for not divining an interpretation of “inability to leave” that petitioner did not clearly advance. On this point, this court reviews denials of withholding of removal “for substantial evidence.” Yali Wang v. Sessions, 861 F.3d 1003, 1007 (9th Cir. 2017) (quotations omitted). “Under the substantial evidence standard, the court upholds the BIA’s determination unless the evidence in the record compels a contrary conclusion.” Arteaga v. Mukasey, 511 F.3d 940, 944 (9th Cir. 2007). Given the (at best) imprecise nature of the phrase “inability to leave” and the record in this case, I cannot conclude that the BIA’s assessment that the proposed social group was effectively defined by the harm lacks substantial evidence.^[5]

Finally, the majority errs in premising its own remand on the BIA’s remand in Matter of A-B-. According to the majority, “the best indication that [mention of physical abuse] does not categorically disqualify [petitioner’s] social group is that the Attorney General remanded Matter of A-B- for the BIA to conduct a proper analysis, . . . [r]ather than simply invalidating th[e] group.” Maj. Op. 35. This is not correct.

5 The majority relies on De Pena-Paniagua v. Barr, 957 F.3d at 94, for the proposition that “[t]here are many reasons a petitioner might be unable to leave a relationship, including a variety of ‘cultural, societal, religious, economic, or other factors.’” Maj. Op. 34 n.11. But for this observation, the First Circuit cited Matter of A-R-C-G-, see De Pena- Paniagua, 957 F.3d at 94, which the Attorney General of course overruled in Matter of A-B-. In all events, De Pena-Paniagua theorized that “a woman’s inability to leave a relationship may be the product of forces other than physical abuse.” Id. at 93. The question in this case is whether the record supports such a characterization of the proposed social group. For the reasons set forth above, the record does not compel the conclusion that the BIA erred in construing the petitioner’s proposed group in the way that it did.

DIAZ-REYNOSO V. BARR 75

The Attorney General in Matter of A-B- was unequivocal that “[h]ad the [BIA] properly analyzed the issues” in Matter of A- R-C-G-, “then it would have been clear that the particular social group was not cognizable.” Matter of A-B-, 27 I. & N. Dec. at 334 (emphasis added). The Attorney General in Matter of A-B- was therefore not somehow suggesting that the BIA in any further proceedings that may occur in Matter of A-B- itself could somehow treat as cognizable the very proposed social group that Matter of A-B- expressly rejected. The standard remand language in Matter of A-B- “for further proceedings consistent with this opinion,” id. at 346, should thus not be construed as a basis for our court to remand a petition where the BIA complied with Matter of A-B-.

* * *

In upholding the Attorney General’s decision in Matter of A-B-, we should have faithfully interpreted it. And in reviewing the BIA’s decision applying Matter of A-B- to petitioner’s application, we should have recognized the BIA’s adherence to an Attorney General decision that is itself based on longstanding precedent and at the very least reasonable. The Attorney General in Matter of A-B- emphasized that under the INA, claims for asylum and withholding of removal based on domestic violence “are unlikely to satisfy the statutory grounds for proving group persecution.” 27 I. & N. Dec. at 320. Today’s decision does not directly quarrel with that understanding, but it sets forth an internally inconsistent analytical framework that is at odds with it, and also at odds with how other circuits have handled this issue.

Where all of this will lead is uncertain. The majority’s opinion remands this case “for further consideration” so the BIA can conduct a “case-specific inquiry” under “the

76 DIAZ-REYNOSO V. BARR

required framework.” Maj. Op. 37, 39, 40. The BIA can be forgiven if it finds this directive unclear. The only certainty is that these issues will be before us again, as the BIA tries to comply with both Matter of A-B- and today’s ruling. I respectfully dissent.