MARCUS, Circuit Judge:

Petitioner Zhou Hua Zhu seeks review of the Board of Immigration Appeals ("BIA") decision denying him asylum and ordering him removed to the People's Republic of China. At issue in this appeal is whether the BIA erred when it overturned the immigration judge's ("IJ") factual findings — particularly the finding that Zhu would likely be forcibly sterilized upon returning to China — not through the prism of clear error review, but rather after its own de novo consideration of the evidence. Under the relevant regulation, 8 C.F.R. § 1003.1(d)(3), the BIA is empowered to review an IJ's factual findings for clear error only. After thorough review, we conclude that the BIA committed legal error by making its own de novo factual findings. We, therefore, vacate the BIA's decision and remand so that the BIA may review the IJ's decision under the proper clear error standard.

I.

A.

Petitioner Zhou Hua Zhu is a native and citizen of the People's Republic of China and was born in Fujian Province. Zhu entered the United States without inspection. On September 8, 1998, the government commenced removal proceedings against him. Initially, Zhu requested asylum and withholding of removal relief due to religious persecution. On October 8, 1998, an IJ denied his application for relief and ordered Zhu removed to China.

Zhu, however, remained in the United States and eventually had three children here. Zhu then submitted a motion to file a successive asylum application. Zhu argued that, as a Chinese citizen who had more than one child, he would be subject to forcible sterilization and a heavy fine if returned to China. Therefore, he was eligible for asylum, withholding of removal, and relief under the United Nations Convention Against Torture ("CAT").

On October 28, 2009, and November 9, 2009, the IJ held hearings on Zhu's application for relief. Zhu testified to the following facts. He is married and has three children. The oldest son, Alvin, was born on December 25, 2002. Alvin suffers from autism and requires personalized treatment. Zhu's other son, Jody, was born on October 26, 2003. His daughter Sherry was born on May 14, 2009. Zhu feared persecution based on China's family planning policy. According to Zhu, Chinese citizens in Fujian with two or more children are sterilized and required to pay a fine. He had heard accounts from relatives and acquaintances who the Chinese government sterilized for having two or more children. His sister-in-law Shen Cho Shin was forcibly sterilized after the birth of her second child. His friend Chen Ca Sin's wife and a cousin's wife, Ling Jing Zhou, were also forcibly sterilized, the latter on June 11, 2009, for having two children. According to Zhu, the family planning policy would apply to him because he remains a Chinese citizen and, under the Chinese Nationality Act, the Chinese Government considers Zhu's American-born children to be Chinese citizens. The Chinese government at one point required his sons to obtain Chinese travel documents when they traveled to China.

Zhu further testified that, after the birth of his daughter, he asked his family to contact the local village committee and ask about the consequences of having three children. In August 2009, the committee delivered a reply letter, which stated that Zhu must be sterilized because he has three children. The letter also required Zhu to report to the family planning bureau within two weeks of returning to China. Zhu presented this letter at the hearing. Zhu also testified that the government levies heavy fines on those who have more than one child. Zhu did not know exactly what the fine would be but estimated it to be 100,000 yuan or more. His earning potential in China would be only 500 yuan a month, which would make it extremely difficult to pay the fine and provide specialized medical and educational care for his eldest son.

Zhu provided documentary evidence establishing the birth of his three children in the United States as well as evidence showing Alvin suffers from autism. Regarding the family planning policy, Zhu submitted individualized evidence including the letter from the local village committee. Zhu also provided other documentary evidence, including letters from individuals in China and official government publications from Fujian Province. In addition, Zhu provided U.S. government publications on country conditions in China, including reports from the State Department and the Congressional-Executive Commission on China. Finally, Zhu provided an affidavit from Dr. Flora Sapio, which disputed information in the State Department's 2007 Profile of Asylum Claims for China.

B.

On November 9, 2009, the IJ granted Zhu's application and found him eligible for asylum based on his well-founded fear of persecution in China. The IJ specifically found Zhu's testimony "credible and consistent." The IJ found that "family planning officials have indicated in records that were submitted to the Court that [Zhu] would be sterilized. This document has not been proven to be incorrect in any meaningful way." The IJ also found that Zhu's American-born children were counted under China's family planning policy, because Zhu adduced evidence that other similarly situated individuals were subject to the family planning policy. The IJ expressed some uncertainty regarding whether the male parent, i.e., Zhu, would be sterilized. However, the IJ stated there was no evidence that the Chinese government would not sterilize the male parent. Thus, the IJ reiterated his findings that Zhu, "if he were to return to China, would be subject to the family planning policy," and that "it appear[ed] reasonable... that the policy would, in fact, be applied to him personally." The IJ then determined that Zhu had demonstrated a well-founded fear of persecution on this ground and granted his application for asylum. The Department of Homeland Security ("DHS") appealed the IJ's decision.

On June 29, 2011, the BIA overturned the IJ's decision based on its determination that "the record does not establish that [Zhu] faces a reasonable possibility of being sterilized or otherwise persecuted for having had children while in the [U.S.]." The BIA cast its decision as a legal determination regarding "whether specific facts are sufficient to meet a legal standard such as a `well-founded fear,'" and therefore proceeded "under de novo review" and with the assumption it could "give different weight to the evidence from that given by the [IJ]." When considering the documentary evidence de novo, the BIA stated it gave less weight to the evidence upon which the IJ had relied and more weight to the State Department's report on country conditions, which did not indicate coercive population control in China and more specifically in Fujian Province. In doing so, the BIA relied on its decision in In re H-L-H- & Z-Y-Z-, 25 I. & N. Dec. 209, 211 (B.I.A.2010), in which the BIA explicitly held that it reviewed de novo the question of whether respondents faced a "reasonable possibility" of future harms such as forcible sterilization.

The BIA specifically contradicted two of the IJ's findings. First, the BIA stated at several junctures that it did not believe Zhu faced "a reasonable possibility of being sterilized." Second, the BIA found that China's family planning policy likely would not count Zhu's American-born children against the one-child limit. Taken together, the BIA denied "there is a policy of forced sterilization of parents who return to China with children who were born outside of that country." Thus, the BIA concluded that Zhu had "failed to carry his burden of establishing he has a well-founded fear of persecution in China based upon the birth of his United States citizen children," and denied his applications for asylum, withholding of removal, and CAT relief. Zhu timely appealed that decision to this Court.

II.

We review "only the BIA's decision," except to the extent that it "expressly adopt[s] the IJ's opinion or reasoning." Imelda v. U.S. Att'y Gen., 611 F.3d 724, 727 (11th Cir.2010). Legal and constitutional questions receive de novo review. Poveda v. U.S. Att'y Gen., 692 F.3d 1168, 1172 (11th Cir.2012). However, this Court must defer to the BIA's permissible construction of ambiguous terms in the Immigration and Nationality Act ("INA") under Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). See Poveda, 692 F.3d at 1176. In addition, the BIA's interpretation of ambiguity in its own regulations warrants deference under Auer v. Robbins, 519 U.S. 452, 117 S.Ct. 905, 137 L.Ed.2d 79 (1997). See Li Shan Chen v. U.S. Att'y Gen., 672 F.3d 961, 965 n. 2 (11th Cir.2011). This Court reviews factual determinations under the substantial evidence test. Thus, this Court must affirm if the BIA's decision "is supported by reasonable, substantial, and probative evidence on the record considered as a whole." Imelda, 611 F.3d at 727 (internal quotation marks omitted). Under this standard, reversal requires finding "that the record not only supports reversal, but compels it." Id. at 728 (quoting Mendoza v. U.S. Att'y Gen., 327 F.3d 1283, 1287 (11th Cir.2003)).

A.

We first consider the government's position that the BIA may review de novo the IJ's finding regarding the likelihood of future events — in this case, the likelihood of forcible sterilization. The government argues such de novo review was proper under the BIA's precedential decisions in In re H-L-H- & Z-Y-Z-, 25 I. & N. Dec. 209 (B.I.A.2010), and In re A-S-B-, 24 I. & N. Dec. 493 (B.I.A.2008), which held that whether an alien has a well-founded fear of persecution — which includes whether it was reasonably possible that a Chinese citizen returned to China would be forcibly sterilized — was not a question of fact but rather one of law or judgment and therefore properly reviewed de novo. See In re H-L-H- & Z-Y-Z-, 25 I. & N. Dec. at 212. The government contends that In re H-L-H- & Z-Y-Z- and In re A-S-B- warrant Auer deference as interpretations of the regulation setting the BIA's standard of review.

The government's position is plainly erroneous and asks us to ignore what we — and common law courts for centuries — have considered to be factual findings. We, therefore, reject the BIA's precedent decisions In re H-L-H- & Z-Y-Z- and In re A-S-B- to the extent they hold that an IJ's finding regarding the likelihood of a future event is not a factual finding, and that the BIA may review such predictive findings de novo. In doing so, we join four other courts of appeals that have confronted and uniformly rejected the BIA's position on this issue.

The pertinent regulation reads:

8 C.F.R. § 1003.1(d)(3). The regulation forbids the BIA from independently engaging in fact-finding and requires it to apply a clear error standard to IJs' factual findings.

The BIA, however, subsequently has interpreted the term "fact" in a cramped fashion. In two decisions on May 8, 2008, the BIA held that "an [IJ's] prediction or finding regarding the likelihood that an alien will be tortured may be reviewed de novo," In re V-K-, 24 I. & N. Dec. 500, 501 (B.I.A.2008), and that the BIA "acted within [its] authority in determining that the [IJ]'s finding of a well-founded fear was not supported by the record" reviewed de novo, In re A-S-B-, 24 I. & N. Dec. at 498. These decisions make clear that, in evaluating either a persecution claim or a torture claim under the Convention Against Torture, the BIA does not consider the likelihood of a future outcome to be a fact pursuant to 8 C.F.R. § 1003.1(d)(3)(i). See In re A-S-B-, 24 I. & N. Dec. at 498 ("This is not fact-finding because, among other reasons, it is impossible to declare as `fact' things that have not yet occurred."); In re V-K-, 24 I. & N. Dec. at 501 (characterizing "a finding as to the degree of possibility of a result occurring" as a "so-called `fact'" not subject to clear error review). Rather, the BIA's decisions suggest, this determination is an indivisible part of the mixed question of law and fact presented by these claims: whether the applicant is likely to be tortured or has a well-founded fear of persecution.

In 2010, the BIA explicitly applied In re A-S-B-'s holding, which addressed well-founded-fear-of-persecution claims generally, to Chinese citizens' claims that they would be forcibly sterilized upon returning to China. See In re H-L-H- & Z-Y-Z-, 25 I. & N. Dec. at 212. The BIA treated the IJ's credibility determination as a factual finding entitled to clear error review, see id. at 211, but reviewed the IJ's forcible sterilization finding de novo because "[d]etermining whether a fear of what may happen in the future is well founded essentially involves predicting future events, and `it is impossible to declare as "fact" things that have not yet occurred.'" Id. at 212 (quoting In re A-S-B-, 24 I. & N. Dec. at 498). Much as it did in this case, the BIA then found, based on its own view of the evidence, that there was not a "reasonable possibility" the petitioner would be forcibly sterilized or required to pay a severe fine. See id. at 213-18.

Because the scope of BIA review is part of the agency's own regulations, the BIA's interpretation of its scope of review is "controlling unless plainly erroneous or inconsistent with the regulation." Auer, 519 U.S. at 461, 117 S.Ct. 905 (internal quotation marks omitted). However, "Auer deference is warranted only when the language of the regulation is ambiguous," Christensen v. Harris Cnty., 529 U.S. 576, 588, 120 S.Ct. 1655, 146 L.Ed.2d 621 (2000), and it cannot shield an agency's attempt "to overcome the regulation's obvious meaning." Id. In this case, the BIA's position is plainly erroneous and inconsistent with the regulation's unambiguous and obvious meaning: the BIA can review the IJ's factual findings, including predictions about the possibility of future events, only for clear error.

What constitutes a fact for 8 C.F.R. § 1003.1(d)(3)(i)'s purposes is identical to what constitutes a fact in federal courts. The language of the regulation and the Attorney General's explanatory comments indicate the intent to emulate, in the relationship between the BIA and IJs, the scope of review that federal appellate courts have over federal district courts. The dichotomy between factual findings reviewed for clear error and legal determinations reviewed de novo is a cornerstone of appellate review in the federal courts. In contrast, this review model did not exist in the immigration system prior to the adoption of § 1003.1(d)(3) on September 25, 2002. See Board of Immigration Appeals: Procedural Reforms To Improve Case Management, 67 Fed.Reg. 54,878, 54,878 (Aug. 26, 2002). With the new regulation, "[t]he Department[ of Justice's] adoption of the `clearly erroneous' standard encompasse[d] the standards now commonly used by the federal courts with respect to appellate court review of findings of fact made by a trial court." 67 Fed.Reg. at 54,890 (citing Dickinson v. Zurko, 527 U.S. 150, 153, 119 S.Ct. 1816, 144 L.Ed.2d 143 (1999)). Indeed, the Department explicitly adopted the standard for clear error drawn from Supreme Court case law. See id. at 54,889 (citing Anderson v. City of Bessemer, 470 U.S. 564, 573, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985)). The consistent intention behind the change was to mimic federal courts' scope of review: "Just as the Supreme Court has concluded that on balance the `clearly erroneous' standard is an effective, reasonable, and efficient standard of appellate review of factual determinations by district courts ..., the Department has concluded that the `clearly erroneous' standard is an effective, reasonable, and efficient standard for appellate administrative review of factual determinations by immigration judges." Id. at 54,889-90.

There is no indication in the regulation's text or the explanatory comments that the term fact has a meaning distinct from its usage in the federal courts. Although the commentary focused on credibility determinations as a special province of the fact-finder, it stressed that "[t]he rationale for changing to a `clearly erroneous' standard of review of fact findings is not limited to the consideration that [IJs] may be better positioned than the Board to decide factual issues, including issues of credibility." Id. at 54,889. Rather, the commentary cited Anderson for the proposition that clear error review also avoids duplication of trial courts' efforts and conserves judicial resources. See id. The regulation's plain intent was that the BIA would conduct its review like a federal court of appeals, and that a fact for BIA purposes would be identical to a fact for federal courts' purposes.

In light of the regulation's text and commentary, the BIA's reasoning in In re A-S-B- and In re H-L-H- & Z-Y-Z- is unsound. The BIA's statement that "it is impossible to declare as `fact' things that have not yet occurred," In re H-L-H- & Z-Y-Z-, 25 I. & N. Dec. at 212 (internal quotation marks omitted), flies in the face of centuries of common-law adjudication. The federal courts long have treated predictions about the likelihood of future events as factual findings to be reviewed for clear error, and the courts have done so in a wide variety of contexts. Perhaps the most obvious example is the calculation of lost future income in a tort action. See Gulf, C. & S.F. Ry. Co. v. Moser, 275 U.S. 133, 135-36, 48 S.Ct. 49, 72 L.Ed. 200 (1927) (reviewing jury instructions regarding lost future income); see also Culver v. Slater Boat Co., 722 F.2d 114, 122 (5th Cir.1983) (en banc) (establishing the proper method for juries as "fact-finders" to adjust loss-of-future-earnings awards to present value). The calculation of future damages requires "estimating the loss of work life resulting from the injury or death, calculating the lost income stream, computing the total damage, and discounting that amount to its present value," Culver, 722 F.2d at 117, and this entire inquiry is the province of the fact-finder. Every factor in the calculation concerns things that have not yet occurred; as Culver noted, "we must remember that the ultimate total damage figure awarded is the sum of a series of predictions, none of which involves mathematical certainty," id. at 121 (internal quotation marks omitted), but nonetheless "these fact-finders must attempt, in some degree, to gauge future events." Id. at 123. Damages for lost earnings are reviewed for clear error. See Nakajima v. United States, 965 F.2d 987, 990-91 (11th Cir.1992).

Calculation of lost earnings is one of many examples of this sort of predictive fact-finding. In tort, awarding future medical expenses demands a very similar inquiry — namely, how much the plaintiff will have to spend in the future to recover from or ameliorate the effects of injuries caused by the defendant. Future medical expenses are also a finding of fact reviewed for clear error only. See Superior Constr. Co. v. Brock, 445 F.3d 1334,1346 (11th Cir.2006) ("[T]he trial court's findings of damages are matters of fact and should be affirmed if not clearly erroneous." (internal quotation marks omitted)); Meader ex rel. Long v. United States, 881 F.2d 1056, 1060 (11th Cir.1989) (same for future medical expenses in Federal Torts Claims Act case).

Contracts law offers another example: a party may receive future lost profits if proven to the jury with reasonable certainty. See Hitchcock v. Anthony, 83 F. 779, 782-84 (6th Cir.1897) (discussing when a jury should be allowed to consider future lost profits as an element of damages); accord Nebula Glass Int'l, Inc. v. Reichhold, Inc., 454 F.3d 1203, 1213 (11th Cir. 2006) ("[S]ince proving lost profits invariably includes some element of prediction about how the market would have behaved but for the defendant's tortious act or breach, Florida courts have often noted that proving lost profits damages is difficult, but by no means impossible."). And in Title VII cases, a victorious plaintiff is entitled to compensatory damages including "future pecuniary losses," not to exceed statutorily established amounts. See 42 U.S.C. § 1981a(b)(3). Notwithstanding the need to make a prediction about the future, the extent of these future pecuniary losses is a question of fact for the jury. See, e.g., Farley v. Nationwide Mut. Ins. Co., 197 F.3d 1322, 1332 & n. 3 (11th Cir.1999).

Indeed, even within the immigration context, we have treated similar predictive findings as factual in nature and therefore reviewed them under the substantial evidence standard. In Arboleda v. U.S. Att'y Gen., 434 F.3d 1220, 1224-26 (11th Cir. 2006), a panel of this Court considered whether the government had established that the petitioner would be able to relocate within Colombia to avoid persecution. The BIA assumed that the Revolutionary Armed Forces of Colombia ("FARC") had persecuted Arboleda in the past but denied him asylum and withholding of removal because the evidence showed Arboleda could relocate to areas of Colombia where the FARC did not operate. Id. at 1222. Because "the record ... compel[led] the conclusion that the FARC operate[d] country-wide in Colombia," id. at 1224, we concluded that the BIA's "finding that the DHS met its burden in establishing that the petitioners could reasonably relocate within Colombia to escape persecution was not supported by substantial evidence in the record," id. at 1226. See also Sepulveda v. U.S. Att'y Gen., 401 F.3d 1226, 1230, 1232 n. 7 (11th Cir.2005) (opining, in dicta, that substantial evidence did not support the IJ's finding that Sepulveda could relocate within Colombia). Like the possibility that a petitioner would be forcibly sterilized, the possibility that a petitioner could relocate safely upon returning to his or her native country is a predictive finding of fact.

Furthermore, in considering our jurisdiction under 8 U.S.C. § 1252(a)(2)(C) and (D) after the REAL ID Act of 2005, we have had occasion to discuss extensively how to distinguish between factual and legal issues amalgamated into a mixed question of law and fact. In an early case interpreting the revised § 1252, which strips federal courts of jurisdiction to review removal orders of alien felons except regarding legal and constitutional questions, the government argued in this Court that a CAT claim "d[id] not raise a `question of law'" but was "actually attempting to challenge a factual determination concerning the likelihood that he will be subjected to torture," Jean-Pierre v. U.S. Att'y Gen., 500 F.3d 1315, 1321 (11th Cir. 2007) — in other words, the exact opposite position to the one it takes in the present case. In Jean-Pierre, this Court squarely rejected that position and held that a CAT claim presented a mixed question of law and fact. Notably, the panel in Jean-Pierre separated the CAT claim into two distinct parts: whether the foreign government would engage in "a particular course of conduct" was a factual issue, which we could not review, see id.; and "[w]hether [that] particular fact pattern amounts to `torture'," which "require[d] a court to apply a legal definition to a set of undisputed or adjudicated historical facts," was a legal issue subject to our scrutiny. Id. at 1322.

Like torture, a well-founded fear of persecution is a mixed question of law and fact. Simply because that mixed question contains an antecedent factual finding, which concerns the future and thus involves some uncertainty, cannot denude such a finding of its factual nature. As we see it, the BIA conflated two distinct parts of the well-founded-fear-of-persecution inquiry: first, determining what has happened to the petitioner in the past and what would occur to the petitioner upon returning to his or her native country; and second, deciding whether what has occurred and what will occur suffice to meet the legal standard for a well-founded fear of persecution. In support, the government relies upon a portion of the regulation's commentary that actually undercuts its argument and presupposes those two parts of the mixed question are separate:

67 Fed.Reg. at 54,890.1 In other words, the commentary envisions the IJ will make a series of factual findings "that form the factual basis for the decision under review" — which the BIA may review only for clear error — and that the final step, determining whether a given set of facts rises to the level of a well-founded fear of persecution, is reviewed de novo.

The finding that a given individual has a certain percent chance of suffering harm like forcible sterilization in the future is not the same as determining that he has a "well-founded fear of future persecution," but rather is part of "the factual basis for the decision under review." The IJ may make that prediction based upon other, subsidiary factual findings — e.g., what China's family planning policy is, and how a petitioner's locality enforces it — and without applying any legal standard. The application of a legal standard to the facts, including those facts which are predictive in nature, occurs at the final step: whether the type of harm and the possibility of it occurring amount to a well-founded fear of persecution. Cf. Jean-Pierre, 500 F.3d at 1322. Distinguishing between these two parts — as we did in Jean-Pierre for torture claims — properly treats the well-founded fear of persecution as a mixed question of law and fact, and divides the factual portion the BIA must review for clear error from the legal portion it may consider de novo.

The BIA has maintained a contrary view about predictive fact-findings despite several adverse decisions from at least four of our sister Circuits. All four courts of appeals have drawn a similar dividing line between fact, even if predictive, and law. The first to address this issue confronted In re V-K-, which held the likelihood of torture was a de novo issue of law or judgment rather than one of fact reviewed for clear error. Kaplun v. Att'y Gen. of the U.S., 602 F.3d 260, 264-65 (3d Cir. 2010). The Third Circuit carefully examined the IJ's three-part determination — (1) the petitioner was likely to come into contact with government authorities; (2) he was likely to be a target for extortion and mistreatment; and thus (3) those harms would rise to the level of torture — and rejected the BIA's argument that this constituted only "a single question, namely, whether there was a clear probability that, if removed, Kaplun would be tortured in the future." Id. at 270-71. Instead, Kaplun parsed the "two distinct parts to the mixed question: (1) what is likely to happen to the petitioner if removed; and (2) does what is likely to happen amount to the legal definition of torture?" Id. at 271. The former is a factual question, which the regulation's language plainly commands the BIA to review for clear error, while the latter is a legal one. Id. The Third Circuit therefore rejected In re V-K- as a plainly erroneous interpretation of 8 C.F.R. § 1003.1(d)(3). In a substantially similar CAT case, the Fourth Circuit also rejected the government's attempt "to recast a finding of likely future mistreatment as something other than a factual finding." Turkson v. Holder, 667 F.3d 523, 529 (4th Cir.2012). Turkson joined Kaplun in holding that "[t]he BIA's decision to subject the IJ's factual findings to de novo review is contrary to the plain language of the governing regulation and is therefore not controlling." Id. at 530. Most recently, the Ninth Circuit joined the Third and Fourth Circuits and admonished the BIA for failing to apply clear error review to the factual portions of a CAT claim. See Ridore v. Holder, 696 F.3d 907, 919 (9th Cir.2012).

Two courts of appeals have confronted practically identical asylum issues to the one presented in this petition. In En Hui Huang v. Att'y Gen. of the U.S., 620 F.3d 372 (3d Cir.2010), the Third Circuit considered an asylum petition from a Chinese citizen who alleged a well-founded fear of persecution under the family planning policy. The petitioner was also a Chinese native from Fujian Province, and she similarly alleged that various relatives from her village had been forcibly sterilized after having multiple children. Id. at 376. The IJ found there was a "strong possibility" Huang or her husband would be forcibly sterilized. Id. at 377. The BIA, much like it did in this case, reviewed that determination de novo and denied that Huang was likely to suffer forcible sterilization. Before the Third Circuit, the government argued that the BIA's precedential decision in In re A-S-B- sanctioned de novo review of an IJ's predictions regarding the possibility of future events, and that In re A-S-B- warranted Auer deference. However, the panel in Huang reiterated the conception of fact the Third Circuit had first explicated in Kaplun: "[T]he probability of an event occurring in the future exists independently of the event itself, and is therefore a separate and distinct `fact' in the relevant legal sense." Id. at 382. The Third Circuit considered Kaplun's reasoning regarding a CAT claim equally applicable to an asylum claim, found the BIA's decision "plainly erroneous," and rejected In re A-S-B-. Id.

The most recent court to consider the matter, the Second Circuit, reviewed a petition from the BIA decision in In re H-L-H- & Z-Y-Z-, which overturned an IJ's decision that the petitioner would be forcibly sterilized and thus had a well-founded fear of persecution. See Hui Lin Huang v. Holder, 677 F.3d 130, 131-33 (2d Cir. 2012). The Second Circuit formulated a similar framework to the one we have adopted today: the IJ "should make clear what it is that the IJ finds is likely to happen to the applicant and how likely it is, those being factual questions," and only then should the IJ "explain the legal conclusion that such treatment of the applicant meets the legal standard for persecution." Id. at 137. The Second Circuit thus rejected In re H-L-H- & Z-Y-Z-.2 In short, the BIA's treatment of an IJ's finding regarding the likelihood of a future event is incompatible with the regulation and its commentary, the fact-law distinctions drawn by this Court in past immigration cases, and the decisions of four other courts of appeals. We therefore reject In re H-L-H- & Z-Y-Z- and In re A-S-B- and hold that an IJ's determination that a petitioner faces a likelihood of being forcibly sterilized in the future is a finding of fact that the BIA may review only for clear error.

B.

The application of the principles we have laid out to the case at hand is simple. The BIA committed legal error by reviewing two of the IJ's factual determinations de novo.

First, the BIA conducted de novo fact-finding — which it is forbidden to do — when it reversed the IJ's determination that there was a likelihood that the local government in Fujian Province would forcibly sterilize Zhu upon his return to China. The BIA specifically cited In re H-L-H- & Z-Y-Z-'s de novo review standard and held "the record does not establish that the respondent faces a reasonable possibility of being sterilized or otherwise persecuted for having had children while in the United States." This determination was fact-finding about the likelihood of a future event, not whether that future event would constitute persecution or whether the likelihood was sufficiently probable to warrant a well-founded fear. And it directly contradicted the IJ's finding that Zhu "would be subject to the family planning policy," and that "it appear[ed] reasonable .... that the policy would, in fact, be applied to him personally." The BIA's decision completely ignored 8 C.F.R. § 1003.1(d)(3)'s command that the BIA "will not engage in factfinding in the course of deciding appeals," § 1003.1(d)(3)(iv), and that the BIA can overturn only "findings of the immigration judge" that "are clearly erroneous." § 1003.1(d)(3)(i).

Based on that error alone we would vacate the BIA's opinion. However, the BIA not only reviewed the IJ's finding that Zhu faced a likelihood of forcible sterilization de novo; it also reviewed one of the factual building blocks of the IJ's prediction, which was a finding of fact about present conditions and circumstances in China, without applying the proper standard of review. In so doing, the BIA conflated subsidiary fact-finding with the ultimate legal determination of whether Zhu had a well-founded fear of persecution. The IJ appears to have found, in substance, that the Chinese government currently treats American-born children of Chinese nationals as Chinese nationals and hence counts such children for purposes of the family planning policy. This finding was a foundational component of the IJ's subsequent prediction: that "the respondent has established that he would be subject to the family planning policy." We emphasize this distinction between the subsidiary finding and the predictive finding, the former of which speaks to present conditions and the latter of which forecasts a future outcome, only for clarity. Of course, both findings are factual, notwithstanding the predictive nature of one, and both should have been reviewed only for clear error. Yet the BIA appears to have rejected the IJ's finding that American-born children currently count for purposes of the family planning policy based primarily on one piece of evidence — an opinion on record at the Law Library of Congress, which provided that a child born in the United States to Chinese parents "is viewed as a U.S. citizen" by China. In the same part of its opinion, the BIA overturned the IJ's predictive finding, stating that "[t]he fact that the child is viewed as a U.S. citizen is a strong indication that the child would not be counted for purposes of the CPC policies." Nowhere in its discussion of either the subsidiary present fact — that the Chinese government considers American-born children of Chinese nationals to be Chinese — or the resulting prediction — that Zhu's children would be counted against the one-child limit, and thus Zhu would be subject to the family planning policy — did the BIA apply clear error review.

We are unpersuaded by the government's suggestion that the BIA never conducted independent fact-finding in this case. The BIA says that it did not find new facts but merely reweighed the evidence before it, a power it locates in In re H-L-H- & Z-Y-Z-, 25 I. & N. Dec. at 212 ("[T]he Board has authority to give different weight to the evidence from that given by the [IJ]."). The BIA's position is incompatible with the language of 8 C.F.R. § 1003.1(d)(3)(i), which states that "[f]acts determined by the immigration judge ... shall be reviewed only to determine whether the findings ... are clearly erroneous." The BIA cannot reverse an IJ's findings and cloak its actions in the euphemistic language of reweighing; accepting the BIA's proposition would essentially permit it to give no weight to the IJ's factual findings, which would cripple the appellate review structure established by § 1003.1(d)(3). To the extent the BIA has the ability to reweigh evidence, this power is limited to reviewing the evidence before the IJ and reweighing it through the prism of clear error review. In other words, the BIA must find that, on balance, the weight of the evidence so strongly militates against the IJ's finding that the BIA "is left with the definite and firm conviction that a mistake has been committed." 67 Fed.Reg. at 54,889. As the regulation's commentary explicitly states, "A factfinding may not be overturned simply because the Board would have weighed the evidence differently or decided the facts differently had it been the factfinder." Id. (citing Anderson, 470 U.S. at 573, 105 S.Ct. 1504). To the extent the BIA suggests it has the power to overturn evidence based on this sort of reweighing, it has abused its discretion and exceeded its authority under 8 C.F.R. § 1003.1(d)(3).

Remand is the appropriate remedy when an administrative agency makes an error of law, for it "afford[s the agency] an opportunity to receive and examine the evidence in light of the correct legal principle." Pollgreen v. Morris, 770 F.2d 1536, 1544 (11th Cir.1985) (citing NLRB v. Enter. Ass'n, Local 638, 429 U.S. 507, 97 S.Ct. 891, 51 L.Ed.2d 1 (1977)). Since the BIA mistakenly reviewed the IJ's findings of fact de novo and conducted impermissible fact-finding, we do not reach the issue of whether substantial evidence supported the BIA's determinations in this case. Accordingly, we grant Zhu's petition, vacate the BIA decision, and remand for the BIA to determine in the first instance whether the IJ's factual findings were clearly erroneous. Only then can the BIA answer the legal question — whether Zhu has established a well-founded fear of future persecution.3

PETITION FOR REVIEW GRANTED; VACATED AND REMANDED.