facts and law in this case. Judge M. Smith wrote that despite the dissent’s claim, the panel’s decision did not invent the cumulative-effect-review requirement; rather that interpretation of the asylum and withholding regulations has been a part of Ninth Circuit and BIA precedent for a quarter century. Judge M. Smith also wrote that it is black-letter law that this court reviews de novo the legal contention that the agency erred by failing to conduct the required cumulative error review, and that where this court finds that the agency erred by failing to consider the cumulative effect, it cannot reach the agency’s bottom-line determination that no past persecution occurred, and instead must remand for the agency to reconsider its determination applying the correct the legal standard. Dissenting from the denial of rehearing en banc, Judge Callahan, joined by Judges Ikuta, R. Nelson, Bumatay, and

** This summary constitutes no part of the opinion of the court. It has been prepared by court staff for the convenience of the reader.

SALGUERO SOSA V. GARLAND 3

VanDyke, wrote that the panel could have held that remand was required because the BIA failed to consider petitioner’s argument that the immigration judge erred by failing to consider his past harm cumulatively. Instead, the majority opinion—based on a misreading of this court’s prior opinions and without consideration of the practical consequences—unnecessarily created a new requirement that when determining whether a petitioner’s past mistreatment rises to the level of persecution, the BIA must apply cumulative-effect review. Judge Callahan wrote that this court’s evaluation of the agency’s analysis of cumulative effect should remain a part of this court’s substantial evidence review, where a lack of analysis of cumulative effects may indicate that the agency’s decision is not supported by substantial evidence in the record. However, there is no support for the proposition that the BIA’s failure to consider the cumulative effect of alleged incidents of past persecution amounts to a legal error. Judge Callahan also reiterated, as noted by Judge Wu in his partial concurrence, that the opinion provides no guidance on how the agency should carry out such a new, mandatory cumulative-effect review. A majority of the panel has voted to DENY Respondent’s petition for panel rehearing only (Dkt. No. 67). Judges S.R. Thomas and M. Smith voted to deny the petition, and Judge Wu voted to grant it. A judge sua sponte requested a vote on whether to rehear the matter en banc, and the matter failed to receive a majority of the votes of the nonrecused active judges in favor of en banc consideration. See Fed. R. App. P. 35(a). The petition for panel rehearing only is DENIED. The sua sponte request for rehearing en banc is also DENIED.

M. SMITH, Circuit Judge, concurring in the denial of

rehearing en banc:

Whenever the issues in a case are fairly presented in the disposition of the original panel, and any dissent from a denial of rehearing en banc merely highlights issues previously presented, it is rarely necessary to prepare a concurrence to an order denying rehearing en banc. In this case, however, even though the government did not request rehearing en banc, one of my colleagues sua sponte did so. The sua sponte call failed to convince a majority of our court and now gives rise to the dissent filed with this order. Because the dissent inaccurately discusses both the facts and the law in this case, this is the rare instance in which I feel compelled to respond in order to obviate confusion on the relevant issues in the future. Since at least 1998, the Board of Immigration Appeals (BIA) and our court has interpreted “past persecution” in the

SALGUERO SOSA V. GARLAND 5

asylum and withholding-of-removal regulations to require cumulative-effect review: that “[t]he key question” when evaluating a claim of past persecution “is whether, looking at the cumulative effect of all the incidents that a petitioner has suffered, the treatment he received rises to the level of persecution.” Sharma v. Garland, 9 F.4th 1052, 1061 (9th Cir. 2021) (quoting Gormley v. Ashcroft, 364 F.3d 1172, 1176–77 (9th Cir. 2004), which in turn is quoting Singh v. INS, 134 F.3d 962, 967 (9th Cir. 1998)). This case presented two narrow, second-order questions regarding what to do in the rare instance where the agency departs from that well- established requirement:

• First, when a petitioner argues that the BIA and Immigration Judge (IJ) failed to conduct cumulative- effect review, is that a contention of legal error subject to de novo review or factual error subject to substantial- evidence review? • Second, if our court determines that the agency failed to conduct cumulative-effect review, do we remand the case back to the agency for it to reconduct its past- persecution analysis with the error corrected, or do we ignore the error and move on to consider the agency’s bottom-line past-persecution finding?

As in all other cases where a petitioner argues that the BIA employed an incorrect legal framework, we held that de novo review applies. Based on the IJ’s plain admission in his decision that he did not conduct cumulative-effect review, we determined that legal error occurred. Complying with the ordinary-remand rule, we did not reach the agency’s bottom-line conclusion that Salguero Sosa failed to establish past persecution. Instead, we remanded the case back to the agency to reconsider Salguero Sosa’s proffered evidence of past persecution while employing the cumulative-effect legal framework that decades of agency and circuit precedent requires—an utterly ordinary immigration/administrative law decision. Nonetheless, after the government chose not to seek rehearing en banc, one of my colleagues sua sponte requested that our court do so. Now after that request failed to convince a majority of our court that rehearing en banc was needed, one of my colleagues pens a dissent that makes much ado about nothing. Try as it might, the dissent cannot conjure a boogeyman from a decision that combined a repeatedly reaffirmed interpretation with basic administrative law principles to grant a narrow remand.

ANALYSIS

I. For Decades, Precedent Has Required Cumulative-

Effect Review

Despite the dissent’s claim, our decision did not invent the cumulative-effect-review requirement. That common- sense interpretation flows from the language of the asylum and withholding regulations and has been a part of Ninth Circuit and BIA precedent for a quarter century. Let’s start with some general principles. When evaluating applications for asylum and withholding of removal, IJs and the BIA determine whether individual applicants have met statutory criteria for relief. By statute, an applicant is eligible for asylum if he or she has “a well- founded fear of persecution” on account of a protected ground. 8 U.S.C. § 1101(a)(42). Also by statute, an applicant is entitled to withholding of removal if his or her

SALGUERO SOSA V. GARLAND 7

“life or freedom would be threatened” on account of a protected ground. Id. § 1231(b)(3)(A). Asylum and withholding regulations interpret that statutory language to mean that an applicant is entitled to a rebuttable presumption of future persecution if he or she has suffered “past persecution.” 8 C.F.R. §§ 208.13(b)(1), 208.16(b)(1). For decades, our court has interpreted this “past persecution” regulatory language to mean that “[t]he key question” when determining past persecution “is whether, looking at the cumulative effect of all the incidents that a petitioner has suffered, the treatment he received rises to the level of persecution.” Sharma, 9 F.4th at 1061; Salguero Sosa v. Garland, 55 F.4th 1213, 1218 (9th Cir. 2022) (collecting cases with cumulative-effect rule statements). Nor is our court alone in interpreting “past persecution” in this manner. The BIA has long asked whether “incidents,” when viewed “[i]n the aggregate, . . . rise to the level of persecution.” Matter of O-Z- & I-Z-, 22 I. & N. Dec. 23, 26 (BIA 1998). Moreover, the UNHCR Handbook—which we look to as “persuasive authority in interpreting the scope of refugee status,” Miguel-Miguel v. Gonzales, 500 F.3d 941, 949 (9th Cir. 2007) (citation omitted)—has this to say on the subject:

[A]n applicant may have been subjected to various measures not in themselves amounting to persecution . . . . In such situations, the various elements involved may, if taken together, produce an effect on the mind of the applicant that can reasonably justify a claim to well‑founded fear of persecution on “cumulative grounds.”

UNHCR, Handbook on Procedures and Criteria for Determining Refugee Status and Guidelines on International Protection (rev. 2019). Indeed, in its petition for panel rehearing only, the government acknowledged that cumulative-effect review is an established “agency interpretation.” Moreover, it is not difficult to imagine how resounding a loss the agency would suffer before a federal court if it reversed course and attempted to interpret the asylum and withholding-of-removal regulations to mandate single-incident review—that no matter how wretched applicants’ lives are, they are not entitled to relief unless they can point to one definitive incident that itself rises to the level of persecution. It is only the dissent that makes the full-throated argument that there is no requirement of cumulative-effect review. To do so, the dissent attempts a sleight of hand with the Korablina cumulative-effect formulation: that “persecution may be found by cumulative, specific instances of violence and harassment.” Korablina v. INS, 158 F.3d 1038, 1044 (9th Cir. 1998). In its view, this “may” language “undermines” there being a requirement of cumulative- effect review because “[m]ay be found” and “must be evaluated” are “not equivalent.” The dissent is correct that “may” and “must” are different words, but it overlooks that “may” and “must” also modify different words in our court’s decisions. The language from Korablina we quote says that “[p]ersecution may be found” through cumulative effects; “must” would make no sense to modify “persecution” because several low-level harms do not amount to persecution, while one instance of serious harm (e.g.,

SALGUERO SOSA V. GARLAND 9

attempted murder) can rise to the level of persecution. By contrast, our opinion says that “incidents must be evaluated cumulatively.” Put otherwise: Our “must” language refers to how past persecution is evaluated, while Korablina’s “may” language refers to whether, when that framework is applied, an applicant’s past experiences actually rise to the level of persecution. By zeroing in on a single linguistic difference, the dissent overlooks important context. The dissent’s exclusive focus on the Korablina formulation also fails to account for the other common formulation of the cumulative-effect requirement used in Sharma and elsewhere: that “[t]he key question” is “whether, looking at the cumulative effect of all the incidents that a petitioner has suffered, the treatment he received rises to the level of persecution.” 9 F.4th at 1061. 1 The cumulative-effect requirement has been in place for a quarter century, and the dissent’s context-devoid reading of two words does nothing to displace it.

1 Dissenting from the panel opinion, Judge Wu did not go quite as far as my dissenting colleague here and suggested only that “there is a question” whether cumulative-effect review is required by precedent. Salguero Sosa, 55 F.4th at 1225 (Wu, J., dissenting in part). The partial dissent quoted three different articulations of cumulative-effect review: (1) that “[w]e look at the totality of the circumstances in deciding whether a finding of persecution is compelled”; (2) that “the cumulative effect of several incidents may constitute persecution”; and (3) that “[t]he key question is whether, looking at the cumulative effect of all the incidents . . . , the treatment . . . rises to the level of persecution.” Id. at 1225. While Judge Wu raised the “question,” he did not provide an alternative reading. And it is unclear to me how else to read these rule statements other than that they stand for the proposition that cumulative- effect review is a necessary part of past-persecution analysis. II. Precedent Requires De Novo Review Followed by a Remand With it established that our precedent requires cumulative-effect review, we can now turn to the only questions actually decided in this case: (A) What standard of review applies to a petitioner’s contention that the agency failed to conduct cumulative-effect review? and (B) What is the remedy? Fortunately, black-letter law supplies the answer to each question. When reviewing agency action, we review contentions of legal error de novo. When we find a legal error, we do not reach the agency’s bottom-line determination (here, that no past persecution occurred); we remand for the agency to reconsider its determination with the legal error corrected. The dissent argues that, by subjecting a legal error to de novo review and then remanding, we found an impermissible “work around” to the substantial-evidence standard of review that applies to our review of the agency’s “administrative findings of fact.” 8 U.S.C. § 1252(b)(4)(B). 2 But employing different standards of review for legal and factual questions is Appellate Review 101. We did not impermissibly “work around” substantial-evidence review; that standard of review (applicable only to factual findings) simply had no bearing on our review of Salguero Sosa’s

2 For purposes of this concurrence, I will assume arguendo that our court has “consistently held that we review a denial of asylum based on a failure to show past persecution for substantial evidence.” But in reality, whether the agency’s past-persecution determination is reviewed de novo or for substantial evidence is the subject of an intra-circuit split. See, e.g., Flores Molina v. Garland, 37 F.4th 626, 633 n.2 (9th Cir. 2022). That intra-circuit split has no bearing on our decision, as we did not actually review the agency’s past-persecution determination—just the legal framework the agency used to make that determination.

SALGUERO SOSA V. GARLAND 11

contention that the agency applied an incorrect legal framework. A. Ninth Circuit Precedent Requires De Novo Review Our court (and every other federal court) reviews de novo whether an administrative agency committed legal error. Let’s return to basic principles. Broadly speaking, petitioners assert two types of substantive errors following the denial of asylum and withholding of removal. The first is legal error: The BIA was mistaken in what it required the petitioner to show. The second is factual error: The BIA was mistaken in concluding that the petitioner failed to make that showing. Different standards of review attend to these different types of error. “We review legal questions raised in a petition for review . . . de novo and any factual findings for substantial evidence.” Tomczyk v. Garland, 25 F.4th 638, 643 (9th Cir. 2022) (en banc). Here, both on appeal to the BIA and in his petition for review before our court, Salguero Sosa contended that the IJ erred in interpreting “past persecution” to require one of his past incidents to itself constitute persecution. See Salguero Sosa, 55 F.4th at 1218, 1220. In other words, he contended that the agency was mistaken in what it required him to show to earn relief. Therefore, we applied de novo review. In doing so, we followed the approach our court has taken with closely analogous contentions. Consider the two examples we offered (id. at 1219): Nexus Framework. Asylum requires a protected ground be “at least one central reason” for persecution, while withholding requires that a protected ground be “a reason” for persecution. Compare 8 U.S.C. § 1158(b)(1)(B)(i) (asylum), with id. § 1231(b)(3)(C) (withholding). Withholding’s “a reason” standard includes “weaker motives.” Garcia v. Wilkinson, 988 F.3d 1136, 1146 (9th Cir. 2021) (citation omitted). In Garcia, the petitioner “contend[ed] that the BIA applied an erroneous legal standard to its nexus analysis for her withholding of removal claim” by applying the heightened asylum standard. Id. We explained that “[w]e review de novo whether the BIA applied the wrong legal standard.” Id. Applying de novo review, we “agree[d]” with the petitioner and “remand[ed] . . . for the BIA to consider whether [the petitioner] is eligible for withholding of removal under the proper ‘a reason’ standard.” Id. at 1146–48. Government-Acquiescence Framework. Article I of the Convention Against Torture (CAT) requires an applicant to show his or her torture would occur with “the consent or acquiescence” of a government official. A previous iteration of the CAT implementing regulations interpreted “acquiescence” to require the government’s actual knowledge of, and consent to, an applicant’s torture. Zheng v. Ashcroft, 332 F.3d 1186, 1194 (9th Cir. 2003). We explained that “[w]e review de novo the BIA’s interpretation of purely legal questions,” subject to applicable deference doctrines like Chevron. Id. at 1193–94. Applying de novo review, we found that the BIA’s interpretation contradicted the “clear intent” of CAT. Id. at 1194. We did not “review the evidence under the correct standard of acquiescence to determine if substantial evidence supports the BIA’s conclusion” and instead “remand[ed] to the BIA to give [it] the first opportunity to apply the correct standard.” Id. at 1196–97.

SALGUERO SOSA V. GARLAND 13

The dissent would have us evaluate an alleged cumulative-effect error “under the existing substantial evidence standard.” However, taking the dissent’s approach would make the cumulative-effect requirement “an outlier in immigration and administrative law.” Salguero Sosa, 55 F.4th at 1219. In Garcia and Zheng, we remanded as soon as we determined there was legal error. The dissent would foreclose that approach here. In the dissent’s view, it is not enough that a petitioner shows that the BIA applied the wrong legal framework, the petitioner must also show that the record compels the conclusion that he or she suffered past persecution. But it offers no good reason why our court should treat a cumulative-effect error “differently than we treat other . . . assertion[s] of legal error.” Id. First, the dissent argues that we cannot rely on the above Garcia and Zheng analogies because each involved a “an interpretation of Congressional intent.” While our rule is rooted in the language of the asylum and withholding regulations (supra section I), such a regulation-versus- statute distinction is immaterial: We interpret both de novo, applying Chevron deference (for statutes) and Kisor deference (for regulations) if applicable. 3

3 The dissent also asserts in one sentence that we imposed a “new, judicially imposed procedural rule.” It does not explain this allegation, but it seems to suggest that our decision violates the Supreme Court’s instruction that we are “‘generally not free to impose’ additional judge- made procedural requirements on agencies.” Garland v. Ming Dai, 141 S. Ct. 1669, 1677 (2021) (quoting Vt. Yankee Nuclear Power Corp. v. Nat. Res. Def. Council, Inc., 435 U.S. 519, 524 (1978)). But we are not imposing a judge-made procedural requirement—like the deemed-true- or-credible rule from Ming Dai, which required IJs to make explicit credibility findings. See 141 S. Ct. at 1677. As explained, cumulative- effect review is an interpretation of what substantive showing the Second, the dissent objects that the cases we cite for the cumulative-effect requirement are cases in which “the IJ did consider cumulative effect[s]” and we reviewed the agency’s no-past-persecution finding for substantial evidence. We recognized as much in our opinion:

Korablina, for instance, took a textbook rule- application-conclusion approach to the issue of whether the petitioner had suffered past persecution. We first described the governing legal rule, Korablina, 158 F.3d at 1044 (“Persecution may be found by cumulative, specific instances of violence and harassment . . . .”), and then applied that rule to Korablina’s testimony, id. at 1044–45 (“Cumulatively, the experiences suffered by Korablina compel the conclusion that she suffered persecution.”). The bottom-line factual conclusion (that substantial evidence did not support the BIA’s finding of no past persecution) necessarily resulted from the application of the legal rule we had stated (that incidents must be evaluated cumulatively).

Salguero Sosa, 55 F.4th at 1219. As we explained, there is nothing untoward with discerning a legal rule from a case that is ultimately decided on substantial-evidence grounds. When we review the BIA’s denial of relief, we (1) assess an applicant’s factual showing against (2) the statutory

regulatory “past persecution” language requires. This is heartland say- what-the-law-is territory for the federal judiciary, not a Ming Dai / Vermont Yankee problem.

SALGUERO SOSA V. GARLAND 15

requirements for relief. When we do (1), we necessarily must explain (2). Our court’s prior statements about cumulative-effect review are not somehow rendered dicta because they preceded substantial-evidence conclusions. Instead, those statements—spread across double-digit cases and over twenty-five years—are better understood as simply meaning what they say: Past-persecution analysis requires cumulative-effect review. References to cumulative-effect review will naturally come in two different types of decisions. In the rare instance like this case where the agency failed to conduct cumulative- effect review, we explain what the agency was obligated to do, find legal error, and remand for reconsideration. In typical cases like those the dissent helpfully collects where the agency properly applied cumulative-effect review, we state the cumulative-effect rule only as a backdrop against which to deferentially evaluate the agency’s past- persecution finding. In either situation, the cumulative- effect rule constitutes a governing legal framework, not merely an element of the agency’s factual analysis. A contention that the agency departed from that framework is an assertion of a legal error that we review de novo. B. Supreme Court Precedent Requires Remand Supreme Court precedent regarding the ordinary-remand rule requires that, after detecting a cumulative-effect error, we remand the case back to the agency. In INS v. Ventura, 537 U.S. 12 (2002) (per curiam), the Supreme Court summarily reversed our court’s decision in which we found error (regarding nexus) and, instead of remanding, went on to reach an issue the BIA had not considered (changed circumstances). Id. at 13–14. The Supreme Court faulted our court for not “respect[ing] the BIA’s role as fact-finder” and held that, “[g]enerally speaking,” a court of appeals should remand after finding error in order to “giv[e] the BIA the opportunity to address the matter in the first instance in light of its own expertise.” Id. at 16–17; see also Gonzales v. Thomas, 547 U.S. 183, 183–85 (2006) (per curiam) (once again summarily reversing our court for violating the ordinary-remand rule). As the “ordinary” modifier suggests, the ordinary- remand rule admits an exception—albeit one that applies “only in narrow circumstances.” Calcutt v. FDIC, 143 S. Ct. 1317, 1321 (2023) (per curiam). Our court need not remand “where ‘[t]here is not the slightest uncertainty as to the outcome’ of the agency’s proceedings on remand” after the error identified by our court has been corrected. Id. at 1321 (quoting NLRB v. Wyman-Gordon Co., 394 U.S. 759, 767 n.6 (1969)). The dissent relies on Chand v. INS, 222 F.3d 1066 (9th Cir. 2000) to argue that, contrary to the ordinary-remand rule, our court should “review[] [a] past persecution finding for substantial evidence even when explicitly recognizing the agency’s failure to analyze cumulative effect[s].” In doing so, the dissent tries to undermine the ordinary-remand rule’s general application by pointing to a case that fell within that rule’s narrow exception. In Chand, the petitioner was an ethnic Indian and practicing Hindu who lived in Fiji during a coup that brought to power ethnic-Fijian nationalists. See id. at 1069-70. In the aftermath of the coup, Chand was beaten three times by the Fijian military, had his home raided and forcibly seized by the Fijian military, and had the Hindu temple he attended burned to the ground. See id. at 1073–76 & n.14. The agency found no past persecution, and Chand filed a petition

SALGUERO SOSA V. GARLAND 17

for review. Our court recognized that “the BIA made no attempt to assess Chand’s claim cumulatively.” Id. at 1074. But given the severity of the harm that Chand had suffered, our court concluded that “[t]here is no doubt” that he suffered past persecution—rendering remand back to the agency unnecessary. Id. at 1076 (emphasis added); accord Calcutt, 143 S. Ct. at 1321 (“not the slightest uncertainty”). Here, we were not presented with an overwhelming record comparable to that in Chand. So, rather than overstep our court’s role, we respected the agency’s role as factfinder and remanded for it to conduct the past-persecution analysis in the first instance while applying the correct legal framework. C. Sister-Circuit Precedent Supports Our Decision Given that our decision is dictated in large part by Supreme Court precedent regarding the ordinary-remand rule, our decision unsurprisingly has the support of our sister circuits. Each circuit that has been presented with the argument that the agency failed to conduct cumulative-effect review has remanded after finding such a failure on the agency’s part. They have not, as the dissent would have it, looked past the agency’s legal error to apply substantial- evidence review to the agency’s bottom-line factual finding that was made pursuant to an incorrect legal framework. The dissent argues that our sister circuits’ precedent is otherwise, and on this point in particular, “I would my horse had the speed of” the dissent’s pen. William Shakespeare, Much Ado About Nothing, act I, sc. 1, ls. 139–40. In just one brief footnote, the dissent asserts that “[m]ost of our sister circuits also appear to evaluate cumulative effect[s] as part of the substantial evidence standard.” But even a cursory read of the dissent’s cited authorities dispels that “appear[ance]” and instead shows that our decision has the unbroken support of our sister circuits. Let’s walk through each of the decisions the dissent cites—and those it conveniently omits: Begin with the sister-circuit decisions that the dissent chose not to include in its footnote. The Second, Fifth, and Seventh Circuits apply de novo review to a petitioner’s contention that the agency failed to conduct cumulative- effect review and then remand if they determine such an error occurred. The Second Circuit faulted the agency for the “the IJ’s apparent (and erroneous) technique of addressing the severity of each event in isolation, without considering its cumulative significance,” and “remand[ed] to the BIA . . . . [to] reconsider the [petitioners’] application in light of the foregoing.” Poradisova v. Gonzales, 420 F.3d 70, 79 (2d Cir. 2005). The Fifth Circuit considered petitioner’s “conten[tion] that the IJ committed legal error by not considering the incidents of harm in the aggregate,” but denied the petition for review because nothing in “the IJ’s decision establishe[d] that the IJ analyzed each incident of harm in isolation.” Eduard v. Ashcroft, 379 F.3d 182, 188 (5th Cir. 2004) (emphasis added). The Seventh Circuit squarely held: “[T]he BIA did not employ the correct standard in evaluating [the petitioner’s] claims. The proper course of action in these circumstances is not for us to decide the question of past persecution in the first instance, but to allow the BIA to re-evaluate the evidence under the proper standard.” Kholyavskiy v. Mukasey, 540 F.3d 555, 571 (7th Cir. 2008). So, three decisions left out of the dissent’s footnote all support the approach we took here. Now, turn to the decisions on which the dissent relies. Like the above, one decision conducts de novo review and remands for the agency to reconsider its past-persecution

SALGUERO SOSA V. GARLAND 19

analysis. The remaining decisions are simply irrelevant to the issue presented in this case: What standard of review applies when a petitioner contends that the agency failed to conduct cumulative-effect review? Mislabeled De Novo Decision. The First Circuit decision cited by the dissent uses substantial-evidence wording, but it ultimately remanded the case to the BIA to “determine whether the harms [the petitioner] suffered . . . meet the standard of past persecution, viewed in the aggregate.” Ordonez-Quino v. Holder, 760 F.3d 80, 92 (1st Cir. 2014). How else is one to understand this holding except that the First Circuit found legal error in the agency’s analysis? In substantial-evidence reversals, the factual issue in question has been definitively established in the petitioner’s favor based on a court of appeals’ determination that “any reasonable adjudicator would be compelled to conclude” that the agency’s finding was erroneous. 8 U.S.C § 1252(b)(4)(B). It would be nonsensical for a court of appeals to remand a case back to the agency for it to reconsider the factual finding that the court already definitively settled. Given this, the First Circuit’s decision can be read only as an imprecisely articulated application of de novo review followed by a remand. Irrelevant Category of Decisions #1. The Third, Fourth, Sixth, and Eighth Circuit decisions cited by the dissent involve petitioners who argued not that the agency failed to conduct cumulative-effect review but that the agency failed to consider a whole category of potentially probative evidence. Fei Mei Cheng v. Att’y Gen., 623 F.3d 175, 193 (3d Cir. 2010) (“the Board failed to consider some of the most serious threats levelled at [the petitioner]”); Baharon v. Holder, 588 F.3d 228, 233 (4th Cir. 2009) (“the IJ and BIA were not then free to base their decision on only isolated snippets of that record while disregarding the rest”); Haider v. Holder, 595 F.3d 276, 288 (6th Cir. 2010) (the IJ “neglected to discuss in his analysis section the repeated stops and searches, the theft of merchandise, or [a] variety of threats”); Ngengwe v. Mukasey, 543 F.3d 1029, 1037 (8th Cir. 2008) (“[t]he IJ did not consider . . . that [the petitioner’s] in-laws confiscated all of her property, and threatened to take her children”). A separate Due Process- based doctrine, which our court already recognizes, governs how to review such an argument. See Cole v. Holder, 659 F.3d 762, 771–72 (9th Cir. 2011) (“[W]here there is any indication that the BIA did not consider all of the evidence before it, a catchall phrase does not suffice, and the decision cannot stand. Such indications include misstating the record and failing to mention highly probative or potentially dispositive evidence.”). Irrelevant Category of Decisions #2. The cited Tenth and Eleventh Circuit decisions do not involve petitioners who contended that the agency failed to conduct cumulative- effect review. Instead, when reviewing agency decisions that correctly conducted cumulative-effect review, those decisions set forth a cumulative-effect rule statement and then evaluated the agency’s no-past-persecution factual finding against that rule. See Ritonga v. Holder, 633 F.3d 971, 975–76 (10th Cir. 2011) (“[w]e do not look at each incident in isolation, but instead consider them collectively, because the cumulative effects of multiple incidents may constitute persecution,” and “[a]fter reviewing the record, we conclude the BIA’s determination . . . is supported by substantial evidence”); Delgado v. Att’y Gen., 487 F.3d 855, 860–62 (11th Cir. 2007) (per curiam) (“[i]n determining whether an alien has suffered past persecution, the IJ must consider the cumulative effects of the incidents,” and “based

SALGUERO SOSA V. GARLAND 21

on the cumulative effect of the two attacks, the continued threatening calls, and the incidents involving the car, Delgado and Ramon have met their burden”). As explained, this category of cases does not in anyway undermine our application of de novo review when presented with a petitioner who does contend that the BIA failed to conduct cumulative-effect review. Supra section I.A. Thus, our sister circuits’ decisions say the opposite of what the dissent suggests they “appear[]” to say. The four circuits that have considered cases similar to this one have taken the same approach: de novo review of the petitioner’s contention of legal error, followed by a remand if such an error is found. III.The Dissent’s Fall-Back Arguments Are Inaccurate Precedent requires cumulative-effect review (supra section I) and that we subject contentions of cumulative- effect error to de novo review (supra section II). Unable to dislodge this precedent, the dissent makes two fallback arguments: that our holding was unnecessary to the disposition, and that it will have negative consequences. The record rebuts both contentions. A. Our Cumulative-Effect Holding Was Necessary to the Disposition Simply put, we needed to explain that remand was required because the government argued that Salguero Sosa failed to identify an error warranting remand. Before our panel, the government argued that remand was inappropriate because, in its view, cumulative-effect review is not a legal requirement but is instead just a framing device a petitioner can use when making a substantial-evidence argument. See Salguero Sosa, 55 F.4th at 1219. By arguing that we “did not even need to discuss legal error” and that our cumulative- effect holding was “totally unnecessary to the disposition of the case,” Judge Callahan’s dissent simply overlooks the arguments with which we were presented in this case. Indeed, it was not until its petition for panel rehearing only that the government made the remand-accepting argument on which the dissent now piggy-backs. By doing so, the government took a heads-I-win, tails-you-lose approach to this case. At the panel stage, the government argued: The court should not remand because cumulative- effect review is not a legal requirement. After we rejected that argument, the government then argued at the panel- rehearing stage: The court’s discussion of cumulative-effect review being a legal requirement was unnecessary—it should just remand. At a minimum, the government forfeited this remand-accepting argument by raising it for the first time in its petition for panel rehearing. See Picazo v. Alameida, 366 F.3d 971, 971–72 (9th Cir. 2004) (order). B. Our Decision’s Impact is Limited Cumulative-effect review has been a part of past- persecution analysis for a quarter century. All we did in this case was remand for further proceedings in the rare instance where the agency deviated from that precedent and practice. Nonetheless, the dissent contends that our decision will have “[c]onsiderable [n]egative [i]mplications” for the BIA and our court. First, the dissent argues that our opinion “offers no guidance for the agency on how to conduct [cumulative- effect] review, or how to indicate it has done so.” But history and the dissent itself indicate that the BIA is perfectly capable of conducting cumulative-effect review. The cumulative-effect requirement goes back to 1998 (supra

SALGUERO SOSA V. GARLAND 23

section I), and the dissent helpfully overviews “multiple examples of where the cumulative effect [of various incidents] provided the basis for a finding of past persecution.” As the dissent notes: “[I]n Korablina, the IJ did consider cumulative effect[s] when evaluating persecution.” Similarly, the dissent observes that in Guo, “the agency actually performed a cumulative-effect analysis.” In the same breadth that the dissent paints cumulative-effect review as an impossible-to-understand requirement, it points to several cases in which the agency applied that requirement without any issue. Moreover, our panel opinion has been on the books now for over half a year—in that time, there has been just one cumulative-effect remand. See Feh v. Garland, 2023 WL 3243987, at *2 (9th Cir. May 4, 2023) (unpublished). Finally, as we emphasized, a petitioner must exhaust a cumulative-effect argument in his or her appeal to the BIA, which gives the BIA an opportunity to correct any defects in the IJ’s analysis or, if it chooses, to affirm on alternative grounds (e.g., a lack of nexus or changed country conditions). See Salguero Sosa, 55 F.4th at 1219 n.1. Second, the dissent laments that we do not address how to resolve a case where “the agency makes some sort of statement that cumulative effects were considered,” and asks “will that [statement] be controlling, or will we be allowed (or even compelled) to look behind that conclusory statement?” It is getting difficult to keep track of whether, in the dissent’s view, we said too much or too little. Compare Dissent section I.B. (criticizing the majority for reaching an issue allegedly “unnecessary to the disposition of this case”), with Dissent section I.C (criticizing the majority for not providing “guidance” on issues not presented). In this case, it was clear from the face of the IJ’s decision that the agency did not conduct cumulative-effect review, as “the IJ expressly stated that he ‘evaluate[d] the nature of each claim the respondent presents in support’ of past persecution and concluded that ‘in every instance what the respondent may have experienced was nothing greater than discrimination focused on him.’” Salguero Sosa, 55 F.4th at 1220. Had we been presented with the situation the dissent postulates, there are analogous doctrines on which we could have drawn. See, e.g., Cole, 659 F.3d at 771–72 (explaining that, where a petitioner argues the agency failed to consider all the relevant evidence, we will look past “a catchall phrase” if the agency “misstat[ed] the record” or “fail[ed] to mention highly probative or potentially dispositive evidence”). Because the dissent’s hypothetical was far afield from this case, we thought it best to leave the task of answering that question to our capable colleagues on future panels who will have the benefit of briefing on the subject. Third, the dissent asserts that we leave it unclear whether our opinion “mean[s] that on appeal we must apply two standards of review—substantial evidence for the agency’s factual determination but de novo review of any allegation the agency did not consider the cumulative effects of the petitioner’s allegations.” Lest there be any confusion, the answer is: yes. It is black-letter law that “[w]e review legal questions raised in a petition for review . . . de novo and any factual findings for substantial evidence.” Tomczyk, 25 F.4th at 643.

CONCLUSION

Our decision followed precedent from the Supreme Court, our court, our sister circuits, and the BIA. We subjected a contention of legal error to de novo review and

SALGUERO SOSA V. GARLAND 25

then remanded the case back to the agency after finding that such an error occurred—a run-of-the-mill immigration/administrative law decision. Following a failed sua sponte call that left a majority of our court unpersuaded, my dissenting colleague simply makes much ado about nothing.

CALLAHAN, Circuit Judge, joined by IKUTA, NELSON, BUMATAY, and VANDYKE, Circuit Judges, dissenting from the denial of rehearing en banc:

We have consistently held that we review a denial of asylum based on a failure to show past persecution for substantial evidence. We must affirm unless our review of the record compels a finding contrary to the agency’s position. The majority opinion in this case seeks to unnecessarily create a new requirement for the BIA based on a misreading of our prior opinions and without consideration of the practical consequences. Instead of simply holding that remand was required because the BIA failed to consider an argument Petitioner raised on appeal, the majority creates a new rule out of whole cloth: “when determining whether a petitioner’s past mistreatment rises to the level of persecution, the BIA must apply cumulative-effect review.” Sosa v. Garland, 55 F.4th 1213, 1218 (9th Cir. 2022) (emphasis added). It further holds that the “BIA’s purported failure to do so is a legal issue we decide de novo.” Id. at 1219. While our prior opinions recognize that past persecution may be found based on cumulative evidence, we have never held that the BIA must apply a cumulative-effect review. This error is unnecessary because (1) the case could have been remanded on the sole basis of the BIA’s failure to address an argument raised on appeal and (2) where the BIA fails to consider the cumulative effect of alleged incidents of past persecution, relief is still available under the existing substantial evidence standard. Moreover, as noted by Judge Wu in his partial concurrence, the opinion provides no guidance on how such a new, mandatory cumulative-effect review is to be carried out. Id. at 1225-26. I. The majority opinion breaks new ground beyond our precedent by framing cumulative-effect review as a legal error instead of evaluating it under the proper substantial evidence review, thereby creating a new, judicially imposed procedural rule governing the analysis of asylum applications. A. The Majority Opinion Incorrectly Applies Our Precedent and Creates a New Legal Requirement For Past Persecution Analysis.

I have no dispute with remanding the case on the basis of the BIA’s failure to address an argument Sosa raised on appeal. Nor do I dispute the proposition that the cumulative effect of multiple past incidents may support or undercut the agency’s determination of past persecution. See Singh v. INS, 94 F.3d 1353, 1358 (9th Cir. 1996) (“[T]he cumulative effect of several incidents may constitute persecution.”). Indeed, the majority opinion cites multiple examples of where the cumulative effect provided the basis for a finding of past persecution. See Sosa, 55 F.4th at 1218 (collecting cases). However, none of those cases supports the proposition that the BIA’s analysis fails as a matter of law

SALGUERO SOSA V. GARLAND 27

when it does not consider the cumulative effect; instead, those opinions are framed in terms of our well-established substantial evidence review. A review of the two principal cases presented by the opinion confirms this point. In Korablina v. INS, petitioner’s evidence of past persecution included a physical attack, her witnessing attacks against her boss who eventually was “disappeared,” threatening phone calls, and the ransacking of her workplace. 158 F.3d 1038, 1044–45 (9th Cir. 1998). The IJ determined that, when considered cumulatively, petitioner’s experience did not amount to persecution. Id. at 1044. We recited the standard of review for upholding BIA determinations, specifically noting the BIA determination must be upheld unless “evidence presented by [the alien] was such that a reasonable factfinder would have to conclude the requisite fear of persecution existed” and that the “alien must demonstrate that ‘the record compels the conclusion’ that she has a well-founded fear of persecution.” Id. at 1041, 1043. In applying that standard, we found that “the credible evidence compels both a finding of past persecution and a well-founded fear of future persecution” and noted that “the IJ’s determination and the BIA’s affirmance that Korablina experienced merely discrimination are not ‘supported by reasonable, substantial, and probative evidence on the record considered as a whole.’” Id. at 1041, 1045. Nothing in Korablina indicates a rule that failure to evaluate cumulative impacts constitutes legal error—instead the case shows that failure to properly consider cumulative effect can result in an agency finding regarding persecution that is not supported by substantial evidence. Indeed, in Korablina the IJ did consider cumulative effect when evaluating persecution. Id. at 1044 (“However, the IJ found, citing Ghaly, that her numerous experiences did not amount to persecution.”). How can a case in which a cumulative- effect analysis was undertaken and then overturned for lack of substantial evidence support the majority opinion’s holding that failure to evaluate cumulative effect constitutes legal error? The majority’s own description of Korablina notes “we held that the BIA’s denial of relief lacked substantial evidence” because of the BIA’s failure to evaluate the cumulative effect of multiple instances of alleged persecution. 55 F.4th at 1218 (emphasis added). Thus, the majority opinion itself recognizes that Korablina did not address an error of law, but a lack of substantial evidence. Id. Guo v. Sessions similarly lacks any requirement of de novo review when looking at an agency’s cumulative-effect analysis. 897 F.3d 1208 (9th Cir. 2018). As with Korablina, the majority opinion inexplicably uses a case in which the agency actually performed a cumulative-effect analysis to bootstrap its holding that a failure to perform such an analysis constitutes legal error. We held in Guo there was a lack of substantial evidence supporting the agency’s denial of relief. Id. at 1213. Petitioner there presented evidence of a police beating, a short detention, and a weekly reporting requirement in support of his past persecution. Id. at 1211. The agency found those actions did not amount to persecution based on a comparison with other cases discussing similar situations. We disagreed, stating that the record as a whole compelled the conclusion that Petitioner suffered past religious persecution. Id. at 1217. There was no discussion of error by the BIA for failing to evaluate the cumulative effect of multiple instances of alleged persecution. In its discussion of the “rule” purportedly set forth in Korablina and Guo, the majority undermines its own

SALGUERO SOSA V. GARLAND 29

position. First, it describes the governing legal rule as “[p]ersecution may be found by cumulative, specific instances of violence and harassment . . .” Sosa, 55 F.4th at 1219. (citing Korablina, emphasis added). Then, in the same paragraph, it describes the legal rule as “incidents must be evaluated cumulatively.” Id. (emphasis added). “May be found” and “must be evaluated” are not equivalent—the first falls within our traditional substantial evidence framework but the latter misstates our precedent to create a new rule of law in the asylum review process. While these cases (and others cited by the majority opinion 1) all stand for the proposition that cumulative effect can be used to support a finding of past persecution (i.e., it may be considered), none of them contains anything close to a statement that a failure to evaluate cumulative effect constitutes a legal error (i.e., it must be considered). Indeed, the majority cites no case mandating that we view a failure to consider cumulative effect in the agency’s persecution analysis as a legal error. Rather, in other cases not discussed by the opinion, we have reviewed the past persecution finding for substantial evidence even when explicitly recognizing the agency’s failure to analyze cumulative

1 See Singh v. INS, 94 F.3d 1353, 1358–59, 1360 (9th Cir. 1996) (noting “the cumulative effect of several incidents may constitute persecution” and lack discussion of any failure to include cumulative-effect analysis); Sharma v. Garland, 9 F.4th 1052, 1057 (9th Cir. 2021) (finding under substantial evidence review that the record did not compel a conclusion of persecution, even when considered cumulatively); Ahmed v. Keisler, 504 F.3d 1183, 1194 (9th Cir. 2007) (stating cumulative effect of harm may be considered in determining whether agency decision is supported by substantial evidence); Krotova v. Gonzales, 416 F.3d 1080 (9th Cir. 2005) (stated cumulative effect of several incidents may constitute persecution and evaluated if agency decision was supported by substantial evidence). effect. See, e.g., Chand v. INS, 222 F.3d 1066, 1074–75 (9th Cir. 2000) (where despite noting “the BIA made no attempt to assess [petitioner’s] claim cumulatively,” we did not remand to the agency for a legal error but instead found under a substantial evidence review that the record compelled a finding of past persecution). 2 Evaluation of the agency’s analysis of cumulative effect should remain a part of our substantial evidence review, where a lack of analysis of cumulative effects may indicate that the agency’s decision is not supported by substantial evidence in the record. Cumulative-effect analysis can support a finding of persecution, but it is not an independent legal requirement imposed on the agency. B. The Majority Reaches an Issue That Is Unnecessary to the Disposition of This Case.

As recognized by the majority opinion and argued by the government’s petition for panel rehearing, Sosa raised a

2 Most of our sister circuits also appear to evaluate cumulative effect as part of the substantial evidence standard. See Ordonez-Quino v. Holder, 760 F.3d 80, 92 (1st Cir. 2014) (determining BIA decision that failed to address cumulative-effect was not supported by substantial evidence in the record); Fei Mei Cheng v. Attorney General of the United States, 623 F.3d 175, 195 (3d Cir. 2010) (same); Baharon v. Holder, 588 F.3d 228, 231, 233 (4th Cir. 2009) (BIA failure to address evidence in the record compelled the court to reverse BIA determination on past persecution); Haider v. Holder, 595 F.3d 276, 287–88 (6th Cir. 2010) (IJ failure to consider relevant facts in the aggregate led to holding that past persecution finding was not supported by substantial evidence); Ngengwe v. Mukasey, 543 F.3d 1029, 1037 (8th Cir. 2008) (remanding a case to the BIA under substantial evidence review); Ritonga v. Holder, 633 F.3d 971, 976 (10th Cir. 2011) (evaluating cumulative effect as part of substantial evidence review); Delgado v. U.S. Attorney General, 487 F.3d 855, 861-62 (11th Cir. 2007) (same).

SALGUERO SOSA V. GARLAND 31

cumulative-effect review argument on appeal and the BIA failed to address it. See Sosa, 55 F.4th at 1220. This is sufficient reason to remand a case to the BIA. Thus, the majority did not even need to discuss legal error for failure to perform cumulative-effect review or suggest any new rule of law; the holding is totally unnecessary to the disposition of this case. Moreover, the holding is unnecessary to our review of asylum cases generally. We have a long-standing standard of substantial evidence review and cumulative effect is properly considered within that framework. In instances where the agency has performed a cumulative-effect review and comes to a conclusion that is not supported by the record, we can find (and have found) that the agency’s decision is in error. The same applies if the agency fails to consider cumulative effect altogether, or only considers some past instances but not others. The majority opinion seems concerned that we would only be able to reverse an agency decision when the record compels a contrary conclusion. But, that is precisely what Congress has deemed our review to be—whether the agency’s decision is supported by substantial evidence. And, that is exactly the way our cases, which the majority cites, deal with this issue. We should not attempt to find a work around of this Congressionally-mandated deferential standard for review of agency decision-making. The majority opinion, however, analogizes its approach to reviewing whether the agency applied an incorrect nexus framework, citing Garcia v. Wilkinson, 988 F.3d 1136, 1146 (9th Cir. 2021), or government acquiescence framework, citing Zheng v. Ashcroft, 332 F.3d 1186, 1194–97 (9th Cir. 2003). Sosa, 55 F.4th at 1219. But our precedent (as well as cases from the Supreme Court) clearly states application of the wrong nexus standard to be a legal error, based on interpretation of the immigration statutes. See Garcia, 988 F.3d at 1146–47. Zheng similarly turns on an interpretation of Congressional intent and the regulations implementing the Convention Against Torture. 322 F.3d at 1194–97. Here, there is no such precedent, and the majority makes no argument that the immigration statutes or regulations mandate the BIA consider cumulative-effect review in any particular way. C. The Majority Opinion Has Considerable Negative Implications Which Are Not Addressed. In addition to erroneously creating a new legal requirement for review of asylum applications when it was unnecessary to do so, the panel majority offers no guidance for the agency on how to conduct such a review, or how to indicate it has done so. As Judge Wu asks in his partial concurrence, what are the practical consequences of this new legal requirement and how shall it be implemented? See Sosa, 55 F.4th at 1225–26 (Wu, J., partial concurrence). Does it mean that in each order denying asylum, the IJ must add the sentence including some sort of magic words to the effect of “the cumulative effect of petitioner’s alleged harms do not rise to the level of persecution?” Or if some particular analysis is required, what does it entail? Does it mean that on appeal we must apply two standards of review— substantial evidence for the agency’s factual determination but de novo review of any allegation the agency did not consider the cumulative effect of the petitioner’s allegations? Moreover, if the agency makes some sort of statement that the cumulative effect was considered, will that be controlling, or will we be allowed (or even compelled) to look behind that conclusory statement?

SALGUERO SOSA V. GARLAND 33

II.

The majority opinion with its lack of direction further muddies the already murky waters of immigration law, creating confusion for immigration judges, petitioners, and practitioners alike and likely leading to inconsistent applications. All for something that is unnecessary and not supported by our precedent—our well-established standard of substantial evidence review is what was required and all that needed to be applied here. For these reasons, I dissent from the denial of rehearing en banc.