dissenting.
I agree with the majority that De Lima’s overbreadth theory is unavailing insofar as he argues that the term “theft *271offense” under the Immigration and Nationality Act (“INA”), 8 U.S.C. § 1101(a)(43)(G), does not include temporary deprivations of property or theft of services. I disagree, however, that we lack jurisdiction to consider his claim that third-degree larceny under Connecticut law doés not qualify as a removable offense under the INA because it includes within its scope at least some forms of fraud that do not satisfy the federal definition of a generic “theft offense.” To the contrary, our precedent on the INA’s exhaustion requirement permits us to address De Lima’s fraud overbreadth claim and achieve the proper resolution of this case.
I.
My colleagues maintain that De Lima’s fraud-based claim is foreclosed by our caselaw because that overbreadth theory was not.argued to the BIA. They rely on Ravindran v. INS, 976 F.2d 754, 760-61 (1st Cir. 1992), where the panel held that it lacked jurisdiction to consider a claim of persecution different from the theory of persecution presented to the BIA. Specifically, we concluded that the petitioner’s assertion to us that he would face persecution on account of his membership in a particular social group was unexhausted because he claimed to the BIA only that he faced persecution on account of his political opinions. Id. at 760.
In applying Ravindran to the very different circumstances of this case, the majority has adopted an unnecessarily restrictive view of the exhaustion requirement. Moreover, where, as here, the petitioner presents a compelling claim on the merits, refusing to consider his challenge to removal is inconsistent with “the fundamental interests at stake.” Gill v. I.N.S., 420 F.3d 82, 87 (2d Cir. 2005). I thus begin with the merits of petitioner’s fraud claim before explaining why we have jurisdiction to provide the relief to which , he is entitled.
A. Connecticut’s Prohibition on Theft by Fraud
As the majority explains, the success of De Lima’s petition for review turns’ on whether his prior conviction under Connecticut law matches up, under the “categorical approach,” with a “theft offense” under the INA, which was the basis of the removal proceedings initiated against him. See 8 U.S.C. § 1101(a)(43)(G). Our inquiry does not focus on De Lima’s specific conduct, but rather, on whether the elements of the state offense of conviction satisfy the elements of the pertinent aggravated felony in the INA list. See Esquivel-Quintana v. Sessions, — U.S. -, 137 S.Ct. 1562, 1568, 198 L.Ed.2d 22 (2017) (“Petitioner’s state conviction is ... an ‘aggravated felony1 under the INA only if the least of the acts criminalized by the state statute falls within the generic federal definition of [a theft offense].”).
A reasonable reading of Connecticut’s third-degree larceny statute, see Conn. Gen. Stat. §§ 53a-119, 53a-124, leaves no doubt that it encompasses at least some crimes that the INA would classify as an “offense that involves fraud or deceit,” under 8 U.S.C. § 1101 (a)(43)(M)(i), and not a “theft' offense” within the meaning of § 1101(a)(43)(G). The BIA has expressly held that the offenses described in these two sections “ordinarily involve distinct crimes,” with a theft offense requiring “the taking of property without consent” and a fraud offense “ordinarily involving] the taking or acquisition of property with consent that has been fraudulently obtained.” Matter of Garcia-Madruga, 24 I. & N. Dec. 436, 440-41 (BIA 2008) (finding that welfare fraud under Rhode Island law is not *272an aggravated felony theft offense as defined in § 1101(a)(43)(G)).
In both its general terms and specific examples, the offense of third-degree larceny in Connecticut includes “the taking or acquisition of property with consent that has been fraudulently obtained.” Id. at 440 (emphasis added). The state’s general larceny definition includes “wrongfully obtaining]” the property of another, without the requirement of lack of consent: •Conn. Geri. St’at. § 53a-119. In addition, the general provision’s list of crimes that constitute larceny includes offenses that fall explicitly on the fraud side of the BIA’s theft-fraud distinction: “[obtaining property by false promise,” id. § 53a-119(3), “defrauding of public community,” id, § ’53a-119(6), and “[a]ir bag fraud,” kl. § 53a-119(16), Although the third-degree larceny statute covers only some of the examples listed in § 53a-119,4 the included crimes are not limited to those that involve the acquisition of property without consent.
Indeed, the government acknowledged at oral argument that Connecticut’s third-degree larceny statute includes crimes within its scope, that, both the BIA and other courts of appeals have characterized as fraud offenses. Yet, government counsel argued—without citation to any authority-—that Connecticut courts have construed fraud in such a way that “fraud” in Connecticut fits,within the generic definition of a theft offense. By way of explanation, she said the Connecticut Supreme Court has, held that consent gained by fraud is not knowing consent. Hence, counsel maintained, a taking by fraud in Connecticut qualifies as a categorical “theft offense” under the INA because it involves “the taking of property without consent.” Garcia-Madruga, 241. & N. Dec. at 440. At a minimum, she suggested, the BIA left open in Garcia-Madruga the question of what constitutes consent, or lack thereof, for purposes of classifying a crime under the INA, and she pointed to the agency’s more recent holding that extortionate takings—which, like fraud, may involve a knowing relinquishment of property—fit the generic definition. See Matter of Ibarra, 26 I. & N. Dec. 809, 811 (BIA 2016) (stating that, because “consent” in extortion offenses is coerced, it “does not constitute the kind of ‘consent’ that exempts an offense from aggravated felony treatment under section 101(a)(43)(G) of the Act”).5 The government appears to maintain that it is up to the BIA,-in each particular instance, to decide whether a state theft crime involves a lack of consent as contemplated by Congress.
This attempt at analytical gymnastics falls flat. Whatever the precise parameters of “consent” within the context of the INA, the term cannot be stretched so broadly as to entirely eliminate the differences between theft and fraud;- By listing the two crimes separately, Congress clearly expressed its view that they are not interchangeable. See Soliman v. Gonzales, 419 F.3d 276, 283 (4th Cir. 2005) (“Where Congress has utilized distinct terms within the *273same statute, the applicable canons of statutory construction require that we endeav- or to give different meanings to those different terms—here ‘fraud’ and ‘theft.’”), Congress, as the BIA has acknowledged, drew the line at consent. See id. at 282 (“The key and controlling distinction between these two crimes is therefore the ‘consent’ element—theft occurs without consent, while fraud occurs with consent that has been unlawfully obtained.” (quoted in Garcia-Madruga, 24 I. & N. Dec. at 439)).
Thus, even if Connecticut chooses to label fraud as a form of theft, that state-law choice cannot override Congress’s judgment to treat the two types of conduct as different crimes for purposes of removal. “The language of a federal statute must be construed to have the meaning intended by Congress, not the [state] legislature.” Drakes v. Zimski, 240 F.3d 246, 248 (3d Cir. 2000); see also Taylor v. United States, 495 U.S. 575, 590, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990) (“It seems to us to be implausible that Congress intended the meaning of ‘burglary’ ... to depend on the definition adopted by .the State of conviction.”); Drakes, 240 F.3d at 248 (“Pronouncing a flower to be.a.rose ... does.not necessarily make it one.”).
I do understand that “consent” is not a fixed concept, and the BIA’s decision in Ibarra illustrates an instance when an intentional relinquishment of property was deemed by the agency to be without consent and, thus, a “theft offense” under § 1101(a)(43)(G). Such- a construction - of the statute, by the-agency charged with administering it, is entitled to deference so long as. it is reasonable. See Chevron, U.S.A. Inc. v. Natural Res. Def. Council, Inc., 467 U.S. 837, 843, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984); Garcia v. Sessions, 856 F.3d 27, 35 (1st Cir. 2017), Thus, if this case turned on whether one particular form of fraud could be construed to include lack of consent as an element, we might need to give the BIA an opportunity to consider the issue before deciding it ourselves.
In this instance, however, the question is whether every crime covered by the Connecticut statute is a removable offense, and the provision plainly sweeps so broadly that we could- not defer to the BIA if'it concluded that all of the conduct criminalized by the provision qualifies as a generic “theft offense” under the INA. See Mellou-li v. Lynch, —- U.S.-, 135 S.Ct. 1980, 1989, 192 L.Ed.2d 60 (2015) (holding that, because the BIA’s reliance on a state drug-paraphernalia offense as a basis for removal under 8 U.S.C. § 1227(a)(2)(B)(i) “makes scant sense,” the BIA approach “is owed no deference under, the doctrine described in Chevron”). Not only would such a determination be inconsistent with the line-Congress has drawn between consensual and non-consensual takings, but the INA also designates fraud as a removable offense only when the. property taken exceeds $10,000 in value. See 8 > U.S.C, § ,1101(a)(43)(M)(i). A conviction for third-degree larceny -in Connecticut will not meet that financial threshold; the offenses covered by § 53a-124 either include -no minimum value for the wrongfully obtained property or specify values of $10,000 or less. See Conn, Gen. Stat. § 53a-124.
Accordingly, there is simply no insight that we could obtain from the BIA relevant to De Lima’s fraud-based over-breadth challenge. To the contrary, it is cle,ar that, if we addressed that claim, the BIA’s order of removal, premised on De Lima’s conviction for third-degree larceny under Connecticut law, could not stand.6 *274The majority resists this conclusion, focusing on the BIA’s use of the word “ordinarily” in Garcia-Madruga. As I have explained, however, it is one thing to say that some fraudulent takings—like extortion— properly may be classified as theft offenses, and an entirely different thing to say that every fraudulent taking proscribed by the Connecticut statute could be so categorized. This statute is inescapably overbroad, and it is simply wrong for the majority to suggest that the BIA might—or could—conclude that it is not. Their rejection of my analysis necessarily rests on their view of the exhaustion requirement, to which I now turn.
B, Exhaustion
Although I understand how my colleagues have read our precedent to foreclose De Lima’s fraud-based overbreadth challenge, their view fails to take into account how the circumstances of this case differ from those underlying most of our exhaustion precedents. The exhaustion requirement only makes sense where, by ignoring it, we would exceed our jurisdiction or violate some principle of administrative law. As I shall explain, neither of those barriers exist here, and the majority’s unreasonably strict application of the exhaustion doctrine unnecessarily produces an unjust result.
As a threshold matter, the majority correctly asserts that “[t]he law is clear that ‘theories not advanced before the BIA may not be surfaced for the first time in a petition for judicial review of the BIA’s final order.’ ” Maj. Op., § C (quoting Pérez Batres v. Lynch, 796 F.3d 157, 160 (1st Cir. 2015)). Indeed, exhaustion is a statutory requirement, see 8 U.S.C. § 1252(d)(1), and, hence, mandatory. See Gill, 420 F.3d at 85. Ordinarily, then, when an alien fails to present an issue to the BIA, we lack jurisdiction to consider it. Mazariegos-Paiz v. Holder, 734 F.3d 57, 62 (1st Cir. 2013). Thus, if De Lima had not presented an overbreadth theory to the agency, I would have to agree with my colleagues that we could not consider that claim. The statute’s overbreadth, however, was the focus of De Lima’s argument to the BIA, and he framed his argument inclusively more than once. See, e.g., Br. to the BIA, at 5 (“The Immigration Judge Erred in Holding that Connecticut Larceny in the Third Degree is an Aggravated Felony”); id. at 7 (“The Connecticut Statute for Third-Degree Larceny is Overbroad and Covers Offenses not Criminalized by-the Federal Definition of Theft”).
The question before is thus only “the level of specificity at which a claim must have been made to have been ‘exhausted’ under § 1252(d)(1).” Gill, 420 F.3d at 85. In my view, the answer to that question must take into account the nature of the claim under scrutiny. After all, the primary purpose of the exhaustion requirement is to protect the agency’s authority over matters within its area of expertise. See, e.g., Mazariegos-Paiz, 734 F.3d at 63 (stating that the exhaustion requirement “afford[s] the parties the full benefit of the agency’s expertise and allow[s] the agency the first opportunity to correct its own bevues”). In determining whether a claim has been properly exhausted, therefore, it is appropriate to consider whether judicial review of that claim would usurp the agency’s function. See id. at 62-63.
With that consideration in mind, the agency is unquestionably entitled to the *275first opportunity to decide issues that depend on facts, particularly when there is an administrative record that includes testimony subject to a credibility assessment. The validity of a petitioner’s claim that he has been, or will be, subject to persecution if removed from the United States is one such issue, and it was the focus in Ravin-dran, the case on which the majority relies to say that we lack jurisdiction over De Lima’s fraud-based overbreadth claim. The panel in Bavindran reasonably rejected the petitioner’s effort to switch theories of persecution, as the facts relevant to his original claim of political persecution could be expected to differ from those relevant to his claimed persecution based on membership in a social group. See 976 F.2d at 761 n.5 (describing at length the multiple factual issues implicated by the unpreserved social-group claim); see also, e.g., Ramirez-Matias v. Holder, 778 F.3d 322, 327 (1st Cir. 2015) (holding that petitioner failed to exhaust claim of persecution based on social group where he did not present that theory to the BIA); Silva v. Gonzales, 463 F.3d 68, 72 (1st Cir. 2006) (holding that petitioner did not exhaust his past persecution claim, explaining that the “narrow argument in his appeal to the BIA is not sufficient to allow a broader inquiry now into the IJ’s factual determination that Silva did not suffer past persecution”).
Here, however, De Lima merely offers an additional legal argument to support his previously made claim that Connecticut’s third-degree larceny provision is overbroad as a matter of law. Significantly, as explained above, we would not be “usurping] the agency’s function” in considering this extension of his claim, as there is only one way to answer the pure question of law that he raises. Mazariegos-Paiz, 734 F.3d at 62. Thus, there is no ambiguity implicating possible deference to the agency’s judgment. See Chevron, 467 U.S. at 842-43, 104 S.Ct. 2778. Moreover, just as importantly, the BIA already has accepted De Lima’s argument, holding in Garda-Madruga that the' taking of property “with consent that has been fraudulently obtained” is ordinarily not a generic “theft offense.” 24 T. & N. Dec. at 440. We do not disturb the “carefully calibrated balance of responsibilities” embodied in the exhaustion requirement where, as here, the petitioner’s reformulated claim involves a’legal issue that the agency previously has resolved. Mazariegos-Paiz, 734 F.3d at 63; see also Dale v. Holder, 610 F.3d 294, 301 (5th Cir. 2010) (noting that “administrative exhaustion requires only that federal courts refrain from addressing an immigration issue until the appropriate administrative authority has had the opportunity to apply its specialized knowledge and experience to the matter” (internal quotation marks and brackets omitted)).
My colleagues confusingly highlight the complexity of the categorical analysis and suggest that, with the approach to exhaustion that I consider appropriate, the BIA and the opposing party would need to anticipate unmade arguments.' That approach, however, does not require thé BIA to address arguments the petitioner did not expressly bring to the agency, and, hence, there is no basis for asserting that it would impose “an unfair and daunting task” on the BIA or- opposing party. The question is whether we treat as exhausted an argument raised to us that is a variation on a legal argument previously raised to the BIA. There is nothing unfair or daunting for the government or for us in addressing such an argument.
How we apply the exhaustion requirement in this case will determine whether De Lima faces mandatory removal based on a clear error of law. Where the purpose of exhaustion will not be advanced and the outcome is so clearly unjust, a strict appli*276cation of that doctrine is both unwise and unnecessary. The Second Circuit has adopted a sensible approach that remains true to the rationale for exhaustion, concluding that the statutory exhaustion provision, § 1252(d)(1), “bars the consideration of. bases of relief that were not raised below, and of general issues that were not raised below, but not of specific, subsidiary legal arguments, or arguments by extension, that were not made below.” Gill, 420 F.3d at 86.
My colleagues suggest that I am forced to rely on Second Circuit law because “[t]here is no precedent in our circuit that even remotely supports defining exhaustion so loosely as to encompass De Lima’s newly minted challenge.” I agree that we have no case adopting the pragmatic, fair approach taken by the Second Circuit. My point, however, is that we have, no law foreclosing such an approach in a case such as. this, where the general claim was raised to the BIA, the dispositive issue is one of law, and there is only one permissible outcome. Moreover, my colleagues incorrectly imply that, in addition to endorsing the Second Circuit’s reading of the statutory exhaustion requirement, I rely on the Gill court’s alternative rationale that courts possess the power to assert jurisdiction to avoid manifest injustice even where there has been a failure to exhaust administrative remedies. See Gill, 420 F.3d at 87.,That reasoning is not part of my analysis, which relies solely on a construction of the exhaustion requirement of § 1252(d)(1)—a task that undeniably belongs to the courts.7
Indeed, where the error is so plain, we should feel obliged to undo the BIA’s incorrect classification of De Lima’s third-degree larceny conviction and the mandatory removal order premised on that finding. The Supreme Court has highlighted the “harsh consequences” that flow from the aggravated felony classification. Moncrieffe v. Holder, 669 U.S. 184, 133 S.Ct. 1678, 1682, 185 L.Ed.2d 727 (2013). Under the INA, the Attorney General may not “grant[ ] discretionary relief from removal to an aggravated felon, no matter how compelling his case.” Id, We should not hesitate to adopt an approach-permitted by our precedent—that would avoid unjustly imposing those consequences. As the Second Circuit observed in GUI, while “there is no jurisdiction of the heart, it does not follow that a court must be completely indifferent to the interests at stake when exercising' lawful discretion or interpreting general statutory language.” 420 F.3d at 87 ii.7 (internal quotation marks omitted).
⅜ * ⅜
In sum, “the merits of [De Lima’s] ... argument are clear-cut in his favor, and (go[] to the very basis for his deportation.’ ” Gill, 420 F.3d at 88 (quoting Marrero Pichardo v. Ashcroft, 374 F.3d 46, 54 (2d Cir. 2004)). We have the authority to hear his claim, and we should do so. I therefore respectfully dissent from the majority’s denial of De Lima’s petition for review.
. The relevant portion of § 53a-124 states that larceny'in the third degree involves the talcing, obtaining, or withholding of motor vehicles valued at $10,000 or less; other property or service exceeding .two thousand dollars in value; certain public records; and specified scientific or technical materials.
. In Ibarra, the immigration judge had found that a conviction under a California penal statute was not categorically an aggravated felony theft offensfe “because the statute proscribes generic extortion in addition to generic theft offenses,” and generic extortion is defined as "obtaining property ‘from another with his consent induced by the wrongful use of force, fear, or threats.’ ” 26 I. & N. Dec. at 810 (quoting United States v. Becerril-Lopez, 541 F.3d 881, 891 (9th Cir. 2008)).
. Indeed, the outcome here is so obvious that there is no need to invoke the well-established *274rule of lenity that, as the majority acknowledges, “plays a role in construing provisions of the INA that trigger deportation or removal.” Maj. Op. at 264. See Fong Haw Tan v. Phelan, 333 U.S. 6, 10, 68 S.Ct. 374, 92 L.Ed. 433 (1948) ("We resolve the doubts in favor of [the alien] because deportation is a drastic measure,...”).
. I am frankly puzzled by my colleagues' reference to the "perfect world" where "Congress might be persuaded” by my analysis. Although Congress has imposed the exhaustion requirement, it is the job of the courts to apply that requirement in a fair and sensible manner.