Chief Judge KATZMANN dissents in a separate opinion.
GERARD E. LYNCH, Circuit Judge: magistrate judge.*152For nearly seventy years, antibiotics have provided dramatic medical advances in the treatment of bacterial infections.1 *153For nearly as long, scientists have been concerned about the problem of antibiotic resistance. Through repeated exposure to antibiotics, some strains of bacteria develop resistance or immunity to particular antibiotics. Such resistance presents a serious threat to human health. Infections in humans caused by antibiotic-resistant bacteria result, on average, in longer hospital stays, worse side effects of treatment, and a greater likelihood of death. In an effort to forestall the development of antibiotic-resistant strains of bacteria, doctors exercise restraint in prescribing antibiotics and are careful to direct patients to use antibiotics only as prescribed.
However, for each dose of antibiotics given to humans for medical purposes, four doses are given to livestock for non-medical reasons to encourage faster, healthier growth. In 2009, 28.8 million pounds of antibiotics were administered to animals raised for food, most of it through animal feed. Unfortunately, research shows that bacteria that develop resistance to antibiotics used in animal feed can transfer to human beings and pose a risk to human health. For that reason, various public-interest organizations have sought to force the Food and Drug Administration (“FDA”) to prohibit the use of certain antibiotics in animal feed. This case arises from one such effort.2
In this lawsuit, the plaintiff organizations contend that the FDA is required by 21 U.S.C. § 360b(e)(l) to proceed with hearings to determine whether to withdraw approval for the use of penicillin and tetracyclines in animal feed, and that the FDA’s denial of two citizen petitions demanding such hearings was arbitrary or capricious within the meaning of 5 U.S.C. § 706(2). The district court accepted plaintiffs’ contention. Because we conclude that plaintiffs and the district court are incorrect, we reverse the judgment of the district court.
BACKGROUND
I. FDA Regulation of Animal Feed Antibiotics
The FDA has statutory authority to regulate new animal drugs3 introduced into interstate commerce. See 21 U.S.C. § 360b(a)(l). New animal drugs are prohibited unless specifically approved by the FDA following a new animal drug application (“NADA”) made by a sponsor, which is usually the drug manufacturer that produced the drug.4 Because antibiotics can be used in animal feed to produce bigger animals that grow faster on less food, many drug manufacturers have sought approval to sell antibiotics for use in animal feed.
In 1951, the FDA approved the first use of antibiotics as ingredients in animal feed to encourage animal growth. Two years later, it approved the first use of antibiotics as drugs in animal feed. But by the late 1960s, the FDA “became concerned about the safety to man and animals of subtherapeutic antibiotic use” both as a general matter and specifically in the context of animal feed.5 See Tetracycline *154(Chlortetraeycline and Oxytetracycline)-Containing Premixes: Opportunity for Hearing, 42 Fed.Reg. 56264, 56266 (Oct. 21, 1977) (“Tetracycline NOOH”). Thus began the decades-long investigation of the danger posed by such use, and the concern about human safety has persisted ever since.
In 1970, prompted by a report published by the United Kingdom’s Joint Committee on the Use of Antibiotics in Animal Husbandry and Veterinary Medicine, the FDA instituted a Task Force to study the problem. In 1972, the Task Force published its report, concluding that: (1) the use of antibiotics in “subtherapeutie amounts” favors the selection of antibiotic-resistant bacteria; (2) animals treated with such doses of antibiotics can serve as hosts for resistant bacteria, which can then be transferred to humans; (3) the prevalence of resistant bacteria had increased; and (4) resistant bacteria had been found in meat and meat products intended for human consumption. The Task Force’s report proposed withdrawing approval for all then-approved subtherapeutie uses of antibiotics unless the manufacturers of the drugs submitted evidence regarding the safety and effectiveness of the drugs as used in animal feed.
In 1977, after receiving the requested information from the drug manufacturers and the recommendation of the Animal Feeds Subcommittee of the National Advisory Food and Drug Committee, the FDA’s Bureau of Veterinary Medicine (“CVM”)6 issued notices of opportunity for hearing (“NOOHs”) with respect to both penicillin and tetracyclines, another family of antibiotics. Penicillin-Containing Premixes: Opportunity for Hearing, 42 Fed. Reg. 43772 (Aug. 30,1977) (“Penicillin NOOH”); Tetracycline NOOH, 42 Fed. Reg. 56264 (Oct. 21,1977). The notices detailed the history of subtherapeutie antibiotic drug use and the scientific data on the safety and effectiveness of such use, and concluded that the drug manufacturers had “failed to resolve the basic safety questions that underlie the subtherapeutie use of [antibiotics] in animal feed.” The Penicillin NOOH went on to state that the Director of the CVM had
conclude[d], on the basis of new information before him with respect to these drug products, evaluated together with the evidence available to him when they were originally approved, that the drug products are not shown to be safe under the conditions of use prescribed, recommended, or suggested in their labeling. The evidence, in fact, indicatefd] that such penicillin use may be unsafe, particularly if the higher or therapeutic levels of penicillin should be used as substitutes for the levels currently used subtherapeutically.
42 Fed.Reg. at 43792. The Tetracycline NOOH stated that the use of such drugs was safe only for a list of specific and strictly limited uses. 42 Fed.Reg. at 56287.
Less than a year after the NOOHs were issued, congressional appropriations committees set aside funds so that the National Academies of Sciences (“NAS”) could conduct further research on the safety and effectiveness of antibiotics in animal feed. The report issued by the House Appropriations Committee included thinly veiled suggestions that the FDA not go forward with the hearing process until the research was completed. See H.R.Rep. No. 95-1290, at 99 (1978). The NAS report, which was largely inconclusive but found that “subtherapeutie use of antimicrobials does *155increase the prevalence of resistance among the E. coli and Salmonella of treated animals,” also recommended that additional studies be conducted. National Academy of Sciences, The Effects on Human Health of Subtherapeutic Use of An-timicrobials in Animal Feed xiv (1980), http://www.nap.edu/catalog.php7record_ id=21.
Two years later, the House committee reiterated its desire to see further evidence before approving the hearing process. A year after that, the Senate Committee on Appropriations noted that the additional studies recommended by the NAS had not yet been conducted and concluded that the “FDA will be expected to continue to hold in abeyance any implementation of its proposal pending the final results of the above research and eviden-tiary hearings.” S.Rep. No. 97-248, at 79 (1981).
In 1981, several industry groups petitioned the FDA to withdraw the 1977 NOOHs. They also sought approval for new uses of antibiotics. On February 1, 1983, the FDA formally denied the petitions. Penicillin and Tetracycline (Chlor-tetracycline and Oxytetracycline) in Animal Feeds; Denial of Petitions, 48 Fed. Reg. 4544, 4556 (Feb. 1, 1983). The published notice accompanying the denials stated that “the Director [of the CVM] does not have any less concern at present about the safety issues that prompted adoption of [the NOOHs]. The Director has not changed his earlier conclusion that the available scientific information warrants the proposed actions.” Id. at 4555. In conclusion, the FDA stated that
[t]he notices of opportunity for hearing represent the Director’s formal position that use of the drugs is not shown to be safe. Therefore, the Director has concluded that he does not wish to withdraw the notices of opportunity for hearing. Instead, the Director wishes to place the notices in abeyance pending completion of the studies mandated by Congress.
The Commissioner [of the FDA] has reviewed the Director’s decision and concurs with it.
That notice was signed by the Commissioner of the FDA.
Meanwhile, several additional studies were either commissioned by various government agencies or conducted by independent multinational organizations. In 1984, the FDA contracted with the Seattle-King County Health Department to conduct yet another study. That study sought to determine how easily antibiotic-resistant bacteria could travel from food animals to humans. It concluded that such transmission was likely. In 1987, the FDA asked the Institute of Medicine (“IOM”) to conduct a review of the risks to human health from subtherapeutic uses of antibiotics in animal feed. IOM found “a considerable body of indirect evidence implicating both subtherapeutic and therapeutic use of antimicrobials as a potential human health hazard,” although it could not establish a definitive direct link. In 1997, the World Health Organization held a meeting of experts to develop a report on the question. The WHO report recommended ceasing subtherapeutic use in animals of any antibiotic that is prescribed for use in humans to combat bacterial infections. Many other reports were also compiled and described in the FDA’s draft Guidance for Industry #209, issued on June 28, 2010.
The FDA never held the hearings it proposed in the 1977 NOOHs. On March 9,1999, a group of public-interest organizations petitioned the FDA, pursuant to § 512(e) of the Food, Drug, and Cosmetic Act (“FDCA”), to withdraw regulatory approval for the subtherapeutic use in animal feed of a specified list of antibiotics, which *156included penicillin and tetracyclines. On April 7, 2005, an overlapping but distinct group of public-interest organizations petitioned the FDA a second time with the same request. Both petitions received preliminary responses, but the FDA issued no final response until after the instant lawsuit was filed.
In the meantime, the FDA issued a series of guidance documents to industry groups, in an effort to implement a voluntary program for gradually reducing the subtherapeutic use of antibiotics in animal feed. The primary mechanism for this hoped-for reduction was an agreement to limit the use of certain antibiotics to therapeutic uses authorized by veterinary prescription. The FDA’s October 23, 2003 Guidance for Industry # 152 detailed the FDA’s conclusions about the dangers posed by subtherapeutic use of antibiotics in animal feed. Guidance 152, by its terms, applied primarily to applications for regulatory approval for new uses of antibiotic drugs. On June 28, 2010, FDA released draft Guidance for Industry # 209, which set out its plan to avoid all uses of antibiotics that were not “judicious.” A disclaimer on Guidance 209 specifies that “[i]t does not create or confer any rights for or on any person and does not operate to bind FDA or the public. You can use an alternative approach if the approach satisfies the requirements of the applicable statutes and regulations.”
At the time this lawsuit was filed, the FDA had issued no final response to the citizen petitions and none of the Guidances discussed above had been finalized.
II. The Instant Lawsuits
Plaintiffs, a group of advocacy organizations,7 filed this lawsuit in the United States District Court for the Southern District of New York on May 25, 2011. They pled two distinct claims. First, they claimed that 21 U.S.C. § 360b(e)(l) compelled the FDA to hold the hearing proposed by the 1977 NOOHs and, if appropriate, withdraw approval for the antibiotic uses the NOOHs listed.8 Second, they claimed that the FDA had unreasonably delayed by failing to respond finally to the 1999 and 2005 citizen petitions, and asked the court to order prompt responses. On July 7, 2011, they filed an amended complaint, which clarified their basis for standing to sue.
On October 6, 2011, after the FDA answered the amended complaint, plaintiffs moved for summary judgment.9 A month later, the FDA issued final responses denying the 1999 and 2005 citizen petitions, effectively mooting the plaintiffs’ second claim. Essentially, the FDA took the position that an alternative strategy for com-batting the ill effects of subtherapeutic use of antibiotics in animal feed would be more efficient than pursuing an effort to withdraw approval for any such uses. By way of explaining its decision, the FDA stated that proceedings to withdraw drug approvals are very costly and lengthy. The FDA also stated that any new proceedings *157would require a new NOOH incorporating new scientific findings on the relationship between human health and subtherapeutic uses of antibiotics in animal feed. Moreover, the FDA argued, it could not grant the petitions because the withdrawal process had to proceed on a drug-by-drug basis. Accordingly, the FDA had decided to pursue an alternative but complementary course of voluntary measures. Shortly thereafter, the FDA formally withdrew the 1977 NOOHs. Withdrawal of Notices of Opportunity for a Hearing: Penicillin and Tetracycline Used in Animal Feed, 76 Fed. Reg. 79697 (Dec. 22, 2011).
In light of these actions, the plaintiffs withdrew their claim to compel action on their petition, which had been mooted by the FDA’s denial, and, on February 1, 2012, filed a supplemental complaint alleging that the denial of their petitions was arbitrary and capricious. The parties then filed renewed cross-motions for summary judgment. The district court ruled separately on the two remaining claims. In a March 22, 2012 order, the district court (Theodore H. Katz, Magistrate Judge) granted plaintiffs’ motion for summary judgment on the NOOH claim.10 Natural Res. Def. Council, Inc. v. FDA (“NRDC /”), 884 F.Supp.2d 127 (S.D.N.Y.2012). The district court ruled that 21 U.S.C. § 360b(e) required the FDA to hold a hearing once it had made a finding that a particular drug use was not safe. It further ruled that the 1977 NOOH constituted or contained such a finding, and that withdrawal of the 1977 NOOH did not effect a withdrawal of that finding. It therefore ordered FDA to institute withdrawal proceedings for the uses discussed in the 1977 NOOH and, unless the manufacturers could rebut the finding, withdraw approval for those drug uses.
In a June 1, 2012 order, the district court (Theodore H. Katz, Magistrate Judge) granted plaintiffs’ motion for summary judgment as to the claims that the denial of the citizen petitions was arbitrary and capricious. Natural Res. Def. Council, Inc. v. FDA(“NRDC II”), 872 F.Supp.2d 318 (S.D.N.Y.2012). According to the district court, the reasons stated in the withdrawal were insufficient to meet even the very limited review authorized by the arbitrary-and-capricious standard. As to the FDA’s claim that withdrawal proceedings are costly and lengthy, the district court ruled that the statute was clear and that these concerns were not relevant. In making this point, the district court relied primarily on Massachusetts v. EPA, 549 U.S. 497, 127 S.Ct. 1438, 167 L.Ed.2d 248 (2007). NRDC II, 872 F.Supp. at 333-34, 337-38. As to the FDA’s claim that it was pursuing alternative voluntary measures to regulate the use of antibiotics, the district court again concluded that the statute was clear and that voluntary measures — effective or not — could not be substituted for the mandatory measures required by the text of the statute.
The government timely appealed both of the district court’s judgments.
DISCUSSION
We review a district court’s decisions on motions for summary judgment de novo. Chandok v. Klessig, 632 F.3d 803, 812 (2d Cir.2011). Summary judgment is appropriate if there is no genuine dispute as to any material fact and the movant is entitled to judgment as a matter of law. Fed. R.Civ.P. 56(a). Here, the facts of the case are undisputed, and the questions posed are purely legal.
I. The Required Hearings Claim
A. The Statutory Text
The principal question presented by this appeal is whether 21 U.S.C. § 360b(e)(l) *158requires the FDA to proceed with withdrawal hearings for certain previously approved subtherapeutic uses of antibiotics in animal feed because the FDA has made a finding that those uses are not shown to be safe for humans. The text of § 360b(e)(l) clearly requires withdrawal of approval once such a finding has been made; it does not equally clearly specify when the agency makes such a finding, and in particular whether the type of finding that mandates withdrawal of approval is a conclusion based on internal agency deliberations that precedes (and then requires) the holding of a hearing, or a finding that represents the conclusion reached as the result of such a hearing.
21 U.S.C. § 360b(e)(l) addresses the FDA’s power to withdraw approval for “new animal drug[s].”11 The text of the statute states that
(1) The Secretary[12] shall, after due notice and opportunity for hearing to the applicant, issue an order withdrawing approval of an application filed pursuant to subsection (b) of this section with respect to any new animal drug if the Secretary finds ...
(B) that new evidence not contained in such application or not available to the Secretary until after such application was approved, or tests by new methods, or tests by methods not deemed reasonably applicable when such application was approved, evaluated together with the evidence available to the Secretary when the application was approved, shows that such drug is not shown to be safe for use under the conditions of use upon the basis of which the application was approved ...;
Id.
The parties dispute the circumstances under which the mandatory language “shall ... issue an order withdrawing approval” comes into play. In particular, they dispute what it means for the Secretary to make a finding, and when that finding occurs. The text makes clear that an order withdrawing approval must be issued (and so far as relevant here may only be issued) upon the occurrence of two conditions precedent — a finding and a hearing. The parties, in effect, dispute the required and anticipated sequence of those conditions.
The government reads the statute as requiring the sequence: hearing, finding, order. In effect, it reads the provision to say, “If, after notice and a hearing, the secretary finds that a drug is not shown to be safe for use,” she is required to withdraw approval of the drug. In this interpretation, the withdrawal process begins with a notice from the FDA to a drug sponsor of its concerns about an drug, and offering the opportunity for a hearing regarding the safety of the animal drug. If, at the conclusion of the hearing, upon consideration of the evidence presented, the secretary finds that the drug is indeed not shown to be safe for use, she must then issue an order withdrawing approval of the drug. That order of events depends upon *159the conclusion that a finding that an animal drug is not shown to be safe can be made only after the drug’s sponsor’s due process rights — notice and an opportunity to be heard — have been respected. Therefore, the mandatory “shall” applies only to the action — withdrawal of approval — that the Secretary must take if the hearing results in a finding adverse to the drug. On the government’s reading, the mandatory “shall” does not apply to the holding of the hearing itself, which the government argues is a discretionary action that the agency may undertake, or not, in its discretion, based on its judgment about whether the scientific evidence and sound public policy warrant instituting proceedings to withdraw approval.
By contrast, plaintiffs favor the sequence: finding, hearing, finding, order. In effect, they read the statute to say, “If the secretary finds a drug is not shown to be safe for use, she shall provide notice to the. applicant, hold a hearing, issue a second finding, and then withdraw approval.” In their interpretation, the initial finding that the drug is not shown to be safe is based on the agency’s internal investigations of the scientific evidence, and comes before any hearing is held. On plaintiffs’ reading, once the agency reaches the conclusion that the drug is not shown to be safe, the mandatory language of the statute becomes applicable — the agency must issue an order of withdrawal, though it must hold a hearing first. The mandatory “shall” thus in effect governs not only the remedy that must follow a formal conclusion after a hearing, but also the process itself; after reaching its initial conclusion that the drug is not shown to be safe, the agency is required to institute proceedings and effectuate them through a hearing, after which (if the evidence present at the hearing sustains the finding) she must issue an order of withdrawal.
As plaintiffs admit, their construction necessarily contemplates two findings that a drug is not shown to be safe for use: one (based on internal deliberations) that triggers the (mandatory) hearing, and another (after the sponsor has been given notice and an opportunity to be heard, and based on the evidence presented at that hearing) that supports the issuance of an order of withdrawal.13 Plaintiffs argue that the initial finding made by the agency is subject to rebuttal by the sponsor at the mandated hearing; in the absence of such rebuttal, the original finding “ripens” and requires issuance of an order of withdrawal.
The text of § 360b(e)(l) itself does not unambiguously express either of these sequences. The syntax makes at least two things clear. First, because the mandatory verb “shall” is linked to “issue an order withdrawing approval” of an NADA if the requisite finding is made, the statute is clear that the withdrawal of the approval is mandatory if the preconditions set in the statute are met. Second, the statute is clear that two such conditions must be met before the requirement that the Secretary “shall” withdraw her approval is triggered: a temporal condition (the withdrawal order may only be issued “after due notice and opportunity for hearing”), and a factual condition (withdrawal is required only “if the Secretary finds ... that such drug is not shown to be safe for use”).
The syntax is not similarly clear as to the temporal relationship between the hearing and the finding, because the *160phrase “after due notice and opportunity for hearing” is inserted somewhat awkwardly between “shall” and “issue.” Different placement of the notice and hearing language could have decisively directed one or the other of the competing interpretations. Had Congress written, “If the Secretary finds [that the drug is not shown to be safe for use], she shall conduct a hearing on due notice to the applicant, and shall withdraw approval if the evidence at the hearing supports the finding,” the plaintiffs would clearly be correct: after making a “finding,” the Secretary would be required to withdraw approval of the drug, but only after a notice and hearing process. In contrast, if Congress had written, “The Secretary shall withdraw approval [of an NADA] if she finds, after due notice and opportunity for hearing to the applicant [that the drug is not shown to be safe],” the FDA’s interpretation would clearly be correct. Unfortunately, it wrote neither, adopting a syntactically awkward variation that leaves the intended sequence ambiguous.
Although the grammar of the sentence as it is actually written does not absolutely compel either reading, we believe that the government’s interpretation is far more plausible, both as a matter of language and as a matter of conventional legal practice.14 As noted above, the plaintiffs’ reading requires not one but two findings, in a sentence that only refers to one. Congress expressly provided that withdrawal of approval is required (indeed, such withdrawal is authorized) only after a hearing is held and a finding is made. The hearing process is thus a critical precondition of the withdrawal order, and as plaintiffs concede, the entire purpose of the hearing is to determine whether the evidence does indeed show that the drug is not shown to be safe for use. The hearing thus eventuates in withdrawal of approval only if the Secretary concludes, based on the evidence of “experience and scientific data” presented at the hearing, that the *161drug is not shown to be safe. Yet according to the plaintiffs’ reading, the statute makes no explicit reference to any such finding at the culmination of the hearing— in plaintiffs’ view, the only finding that Congress expressly requires, and the one that the grammatical construction of the sentence makes prerequisite to the withdrawal of approval, is the finding that the Secretary makes before the hearing even takes place.15
Similarly, while the plaintiffs’ reading would make the initial internal administrative finding of a lack of showing of safety the trigger for a mandatory hearing, the statute does not grammatically link the only “finding” referred to in the statute to a mandatory hearing, but rather to a mandatory withdrawal of approval. Moreover, the statute does not require withdrawal of approval based solely upon an internal, pre-hearing finding — withdrawal of approval must await the conclusion of the hearing, at which further findings would have to be made. At that point, the withdrawal is no longer the mandatory consequence of the initial finding — if the hearing demonstrated the safety of the drug, withdrawal of approval would not be required, or even permitted. It is, instead, the consequence of the further finding at the end of the hearing, based on the evidence presented there. According to plaintiffs, Congress meant to mandate that upon making the “finding” referred to in § 3609(e)(1)(A), the Secretary is required to hold a hearing, and yet Congress provided that upon making such a finding the Secretary “shall issue” not a notice of opportunity for a hearing, but a withdrawal of approval of the drug. In short, it would be singularly odd for Congress to have chosen the language that it did to describe the process that even the plaintiffs concede it intended.
The government’s preferred reading yields no such difficulties. While it is true that the statute would read more smoothly, and would more clearly express the government’s position, if the phrase “after due notice and an opportunity for hearing” were placed after “finds,” rather than between “shall” and “issue,” nothing in the statutory language needs to be twisted to yield the government’s interpretation. Although the placement of the notice and hearing provision is awkward (on either side’s interpretation), even as placed, it is entirely consonant with the government’s reading. There is nothing syntactically difficult or odd about providing that the Secretary shall withdraw her approval of a drug, after a notice and hearing process, if a finding is made (after such process) that the drug is not shown to be safe.
B. Context
The parties call our attention to various aspects of the larger statutory context that might cast further light on the meaning of this particular provision. The cited portions of the statute, however, do not provide much help in clarifying the meaning of the provision in question, and certainly *162do not provide sufficient instruction to overcome the reading derived from the language of the debated text itself.
As the parties note, different language within the same statutory subsection provides for emergency withdrawal of approval for animal drugs “if the Secretary ... finds” that the drugs pose an “imminent hazard to the health of man or of the animals.” 21 U.S.C. § 360b(e)(l) (last paragraph). In such a case, the Secretary — but not any delegate — may immediately and without a hearing suspend approval for the drug in question. That provision can be used to support either side. On the one hand, plaintiffs argue that its language supports the notion that the Secretary can make “findings” by an internal administrative process, without notice or a hearing. On the other hand, the government argues that the special exception permitting emergency interim relief to prevent “imminent hazards,” and the reservation of authority to make such emergency findings to the Secretary, serves to underscore the general and otherwise-applicable rule that findings that induce final agency actions adverse to applicants must be made after notice and hearing, and must represent the final conclusion of the agency, rather than an interim judgment delegable to lower-ranking officials. Neither of these inferences from the emergency suspension provision can be dismissed as implausible, but neither persuasively illuminates the process anticipated by the language in question in this case.
The district court also relied on a distinction between § 360b(e)(l) and the preceding subsection of the statute, § 360b(d)(l), to support its conclusion that findings could precede hearings for purposes of § 360b(e)(l). Section 360b(d)(l) — which lists the permissible grounds on which the Secretary may initially deny an application for approval of a new animal drug — clearly states that findings of fact must occur after a hearing, by placing the “after due notice [and] hearing” language immediately after the verb “finds.”16 That difference in language, the district court concluded, suggests that Congress intended different schemes under the two subsections. The government challenges that reading, however, arguing that § 360b(d) is expressly connected to the language of the preceding § 360b(c)(l), which in turn explicitly contemplates that some findings may be made without a hearing.17 In that context, it makes sense that Congress would use more sharply contrasting language to distinguish the findings made under subsection (d) from those required in subsection (c). By contrast, under § 360b(e)(l), no order withdrawing approval may issue before a hearing unless the drug represents an “imminent hazard.” In the government’s reading, the difference in language between subsections (d) and (e) reflects the differing procedures for approval of a new drug and for the withdrawal of approval of a previously approved drug, rather than any intention to limit agency discretion to institute, pursue, or abandon procedures to withdraw drug approvals.
*163Again, we find both parties’ inferences from the language of § 360b(d) reasonable. But neither is sufficiently compelling either to strongly corroborate or to seriously undermine our reading of the text. As noted above, it is unquestionably clear from the text that the mandate to order withdrawal only applies after the agency has held a hearing. Indeed, it is clear from the text that an order withdrawing approval may not be entered (except in the emergency circumstances referred to in § 360b(e)(l)) without providing notice and a hearing to the drug’s sponsor. It seems to us that, when a statute provides that an agency must take some action after a hearing “if it finds” something to be true, the more persuasive reading is that the finding referred to is the fruit of the required hearing.
C. The Relevant Regulations
Both parties argue that various regulations implementing the statute support their respective interpretations. For the reasons set forth below, we do not find these arguments especially helpful.
The government argues that the FDA’s interpretation of the statute is entitled to deference. We generally give deference to an agency’s interpretation of statutes that the agency administers. See Chevron U.S.A., Inc. v. Natural Res. Def. Council, Inc., 467 U.S. 837, 844-45, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). The Supreme Court has held that the FDA is entitled to deference when it interprets Title 21 of the United States Code, FDA v. Brown & Williamson Tobacco Corp., 529 U.S. 120, 132, 120 S.Ct. 1291, 146 L.Ed.2d 121 (2000), because the FDA, as designee of the Department of Health and Human Services, is the agency tasked with administering the FDCA. We will therefore defer to the FDA’s interpretation if Congress has not directly spoken to the precise question at issue and the agency can point to an official interpretation that sets forth a permissible construction of the statute.
However, before we defer to an agency’s interpretation of a statute, we must identify an agency document setting forth that interpretation. The level of deference to an agency’s interpretation of its own statute depends on the nature of the document setting forth the interpretation. Regulations promulgated after a period of notice and comment are typically granted the relatively strong form of deference described by Chevron. See United States v. Mead Corp., 533 U.S. 218, 228-30, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001). We give substantially less deference to post hoc interpretations offered only for purposes of litigation, particularly when those interpretations represent a “departure from prior norms.” See Am. Fed’n of State, County, & Mun. Emps. v. Am. Int’l Grp., Inc., 462 F.3d 121, 129 (2d Cir.2006), quoting Atchison, Topeka & Santa Fe Ry. Co. v. Wichita Bd. of Trade, 412 U.S. 800, 808, 93 S.Ct. 2367, 37 L.Ed.2d 350 (1973). The government urges us to apply Chevron deference to the FDA’s interpretation of the statute as embodied in its notice and comment regulations.
The government, however, overlooks a basic predicate of administrative deference. In order to merit deference on a given issue, a particular regulation must shed light one way or the other on the issue. The government points to three regulations to support its position, but none of them help decide the question before us: whether the findings referred to in § 360b(e) precede hearings, or follow them. The government concedes that the regulations it cites do not explicitly undertake to interpret the statutory provision at issue and answer the question before us, but nevertheless argues that the regulations presuppose an answer to that question.
*164First, the government relies on 21 C.F.R. § 5.84, which authorizes the Director of CVM, as the Secretary’s delegate, to issue NOOHs on the latter’s behalf.18 Specifically, the regulation provides that the Director may issue NOOHs or, if the sponsor has waived the right to a hearing, the Director may issue an order of withdrawal. But that regulation fails to give any indication about what, if any, conditions might require a hearing. The government argues that, because § 5.84 represents a partial delegation of the Secretary’s duties under § 360b(e)(l) only for purposes of issuing NOOHs, it does not authorize the Director to make “findings,” and that therefore the issuance of an NOOH is never preceded by a finding as defined in the statute. But the regulation merely states that “[t]he Director and Deputy Director [of the CVM] are authorized to issue [NOOHs] ... and to issue notices of withdrawal of approval when opportunity for hearing has been waived.” 21 C.F.R. § 5.84(a)(1) — (2). It is thus equally plausible to read § 5.84 as delegating to the Director of CVM19 the responsibility for making any findings that might trigger the mandatory issuance of NOOHs and any resulting actions.
The government next points to 21 C.F.R. § 514.200(c), which sets out the possible responses a sponsor may make to an NOOH, and describes the requisite showing a sponsor must make to secure an actual hearing as opposed to a decision on the papers. It argues that that regulation precludes plaintiffs’ reading because any pre-hearing findings are necessarily preliminary. Specifically, the regulation requires a sponsor seeking a hearing to
giv[e] the reason why the application should not be refused or should not be withdrawn, together with a well-organized and full-factual analysis of the clinical and other investigational data he is prepared to prove in support of his opposition to the Commissioner’s proposal. A request for a hearing may not rest upon mere allegations or denials, but must set forth specific facts showing there is a genuine and substantial issue of fact that requires a hearing. When it clearly appears from the data in the application and from the reasons and a factual analysis in the request for the hearing that no genuine and substantial issue of fact precludes the refusal to approve the application or the withdrawal of approval of the application (for example, no adequate and well-controlled clinical investigations to support the claims of effectiveness have been identified), the Commissioner will enter *165an order on this data, stating his findings and conclusions.
21 C.F.R. § 514.200(c) (emphasis added). The government argues that, because a decision to grant a hearing represents only the Commissioner’s determination that there may be a “genuine and substantial issue of fact precluding] ... the withdrawal of approval of the application,” it cannot represent a conclusive finding triggering a mandatory duty. By contrast, the plaintiffs argue, and the district court agreed, that not only does § 514.200(c) fail to support the position that findings may take place only after a hearing, but it also explicitly contemplates findings even in the absence of a hearing. If the sponsor’s written response to a NOOH is insufficient, the Commissioner is to enter an order of withdrawal based on, among other things, the data in the initial application.
The regulation, it seems to us, simply provides a mechanism for a kind of “summary judgment” proceeding that might obviate an evidentiary hearing. If the applicant requests a hearing, the decision to withdraw approval of the drug must be based either on formal “findings” derived from the evidence adduced at a hearing or, where that the request for a hearing does not raise a genuine issue of disputed fact, on a summary judgment-like conclusion. To that extent, the regulation is consistent with the government’s basic contention that “findings” normally result from an adjudicative process, and are made after that process is instituted. It does not, however, directly address — let alone lead us to conclusively reject — plaintiffs’ contention that the entire proceeding is triggered by an agency “finding.” We therefore conclude that this regulation does not embody an unambiguous interpretation of § 360b(e)(l) to which we must defer in this case.
Finally, the government urges us to take account of 21 C.F.R. § 10.55(b)(2)(i), which provides for the separation of investigative and adjudicative responsibilities within the FDA in the event of a hearing. Specifically, that regulation provides that, from the time of the announcement of a formal hearing, CVM will be “responsible for all investigative functions” and for presenting the FDA’s case before the adjudicator. According to the government, this separation of functions “reflects FDA’s understanding that throughout the withdrawal process, CVM does not speak on behalf of’ the FDA. Because CVM cannot speak on behalf of the Commissioner once hearing proceedings have been instituted, the government argues that any “finding” by CVM cannot trigger a mandatory duty on the part of the Commissioner. That argument has one major flaw: the finding that plaintiffs argue triggered the FDA’s duty to proceed with the hearing necessarily preceded the period of separation of administrative functions. Because functions are separated only upon publication of an NOOH, anything that precedes or is included in the NOOH might have represented an action by the Commissioner through her delegate, the Director of CVM.
We are therefore unable to identify a regulation promulgated by FDA pursuant to its notice and comment rulemaking authority that clearly reflects a definitive interpretation of § 360b(e)(l). While the regulations relied upon by the FDA do not expressly adopt or unambiguously require any particular interpretation of the contested language to which we must accord Chevron deference, they still provide some guidance. As we discuss below, we believe that the regulations relied upon by the FDA reflect a conventional understanding of the administrative process that is consistent with the interpretation of § 360b(e)(l) advanced by the government. We cannot conclude, however, that any of those regulations directly speak to the spe-*166cifíc question of statutory interpretation before us, or reflect a clear adoption by the agency of any position on that question.
Plaintiffs, for their part, also seek support in FDA regulations. They argue that the regulation most on point is 21 C.F.R. § 514.115(b)(3)(ii), which details the procedures for withdrawal of approval of an NADA. That regulation provides: “The Commissioner shall notify in writing the person holding an application approved pursuant to section 512(c) of the act and afford an opportunity for a hearing on a proposal to withdraw approval of such application if he finds ” that one of the conditions described in § 360b is met. 21 C.F.R. § 514.115(b)(3) (emphasis added). That regulation unquestionably lends plausibility to plaintiffs’ reading of the statute because it clearly contemplates that the Commissioner must make some sort of finding before the issuance of an NOOH. Indeed it was largely on the basis of § 514.115(b)(3)(ii) that the district court interpreted the statute to require the FDA to proceed with the hearing.
The government, however, argues that this regulation is inapposite because “identical words need not have identical meanings when used in different contexts” and the FDA’s use of “finds” in the regulation refers only to a preliminary finding. While the government’s argument is hardly compelling, the regulation can indeed be read as it proposes. Moreover, that the FDA regulation refers to a finding made by the Commissioner before instituting a hearing does not mean that the regulation is intended to set forth the agency’s definitive interpretation, or indeed any interpretation at all, of the statutory language whose meaning the parties here dispute. In other words, plaintiffs’ reliance on § 514.114(b)(8)(ii) suffers from the same flaw as the FDA’s reliance on the regulations that it cites. All of the cited regulations are drafted to define administrative procedures, and not to interpret the mandate set forth by Congress in § 360b(e)(l). The use of certain language in those regulations or the nature of the procedures that they create may lend some support to the position of one side or the other, but the regulations simply cannot be said to answer the question before us.
In short, we are not required to defer to an agency’s interpretation of a statute when its regulations do not directly address the question before the Court, and when the language of one regulation appears to be in tension with the agency’s interpretation of the statute advanced for purposes of litigation. We therefore conclude that Chevron deference does not provide an answer to the question before us.
D. Background Legal Concepts
We take some comfort from the fact that our interpretation of the statutory text is consistent with ordinary understandings of administrative and judicial litigation processes. In interpreting a statute, courts generally presume that Congress acts “against the background of our traditional legal concepts.” United States v. U.S. Gypsum Co., 438 U.S. 422, 437, 98 S.Ct. 2864, 57 L.Ed.2d 854 (1978). See also United States v. Pacheco, 225 F.3d 148, 157 (2d Cir.2000). Of course, Congress may depart from such traditions; it may use words in ways that are unconventional, or adopt innovative procedures. But when a statute does not provide clear direction, it is more likely that Congress was adopting, rather than departing from, established assumptions about how our legal or administrative system works. We will not lightly assume a less conventional meaning absent a clear indication that such a meaning was intended. In our view, the interpretation advanced by the government is more in accordance with such conventions.
*167First, the government’s interpretation is more consistent with our usual understanding of an administrative “finding.” An agency “finding” typically represents an official determination, reflecting a final, deliberative decision issuing at the conclusion of a process, and taking a fixed form embodied in an identifiable document. Judicial or administrative findings most commonly are adopted not as a prerequisite but as a consequence of a hearing or other official proceeding. For example, Black’s Law Dictionary defines the verb “find” as “To determine a fact in dispute by verdict or decision, ” Black’s Law Dictionary 707 (9th ed.2009) (emphasis added), and the noun “finding of fact” as “A determination by a judge, jury, or administrative agency of a fact supported by the evidence in the record, usu. presented at the trial or hearing, ” id. at 708 (emphasis added). Thus a “finding” traditionally occurs after adversarial parties are given notice of a hearing and an opportunity to be heard there, at least if hearings are contemplated as part of the administrative scheme.
Other areas of the law define “findings” to mean written conclusions issued only at the completion of an administrative process. For example, for purposes of judicial review of agency adjudication under the Social Security Act, we have held that the term “findings” refers to the agency’s on-the-record determinations at a hearing, and that the agency has an affirmative duty to develop the record during the hearing to facilitate judicial review. See Pratts v. Chater, 94 F.3d 34, 37 (2d Cir.1996); see also 42 U.S.C. § 405(g). Similarly, albeit in a context that does not necessarily contemplate hearings, Federal Rule of Evidence 803(8)(A)(iii), which provides an exception to the hearsay rule for written public records and reports, permits admission into evidence “findings from a legally authorized investigation,” and the Supreme Court has interpreted the rule as requiring a “conclusion ... based on a factual investigation.” Beech Aircraft Corp. v. Rainey, 488 U.S. 153, 170, 109 S.Ct. 439, 102 L.Ed.2d 445 (1988) (emphasis added). In these instances, “findings” are written and issue only at the conclusion of the entire process.
We do not suggest that the term cannot be, or indeed is not, sometimes used in a different sense. As plaintiffs point out, even in this very regulatory scheme, the FDA uses the word “find” in a context that clearly refers to a preliminary finding that does not share the characteristics discussed above. See 21 C.F.R. § 514.115(b)(3)(ii), discussed above. Similarly, the emergency suspension proceedings in § 360b(e)(l) itself refer to findings that are made by the Secretary without a hearing, though, notably in that context, the finding is the basis for an action that has an immediate legal effect on the rights of a drug sponsor, rather than being the basis of a decision to institute a process that may eventually lead to such an effect. Nevertheless, where the context does not clearly indicate to the contrary, typical usage suggests that an administrative finding reflects the agency’s final decision issued at the conclusion of a process, rather than a preliminary assessment that contemplates further proceedings before final action is taken.20
Second, the function of the finding contemplated by § 360b(e)(l), and the man*168date that Congress attached to the making of such a finding, is consistent with this more natural meaning. As plaintiffs themselves emphasize, a finding by the Commissioner that a drug is not shown to be safe requires the FDA to “issue an order withdrawing approval” of the drug. 21 U.S.C. § 360b(e)(l). The FDA is not accorded discretion to adopt a different remedy. The consequences of such a withdrawal are significant for society and for the sponsor or manufacturer of the drug. It is logical to assume that Congress would mandate withdrawal of approval of a drug when it has been determined that the drug has not been shown to be safe by a formal decision of the agency, after a careful hearing at which evidence both for and against the safety of the drug has been presented. In contrast, it would seem peculiar for Congress (absent an emergency of the sort authorizing unilateral agency action provided for in the last paragraph of § 360b(e)(l)) to mandate such a strong remedy based not on a final decision by the agency head after a full deliberative process, but on a preliminary conclusion reached by lower-level officials before those affected have had an opportunity to be heard.
And of course, as plaintiffs in effect concede, Congress clearly has not done so. While the language of the statute dictates that withdrawal of approval is the necessary consequence of a finding that an animal drug is not shown to be safe, the statute requires notice to the sponsor and an opportunity for a hearing before a final order of withdrawal may issue, and plaintiffs agree that such an order may issue only if the hearing results in a finding by the Secretary that the drug is not shown to be safe. Under the literal words of the statute, as well as in accordance with common sense, the agency must issue an order withdrawing approval when it finds that the drug is not shown to be safe — something that, as a matter of statutory command and due process, may only occur after the hearing.
Third, the administrative process contemplated by the government’s interpretation of the statute accords with our traditional expectations of governmental enforcement of legal rules. The traditional model of administrative or judicial enforcement features an investigation by executive or administrative personnel, followed by the issuance of a case-initiating document that sets forth the conclusions or charges reached by the prosecuting authority, followed by a hearing. That process culminates in a final adjudication that is reached by the agency and embodied in a formal decision, and imposes whatever remedies or penalties are applicable. In civil and criminal actions, the initial conclusions of the administrative agency or executive officer that lead to the fifing of a lawsuit and an adjudication by a court are not thought of as “findings” and do not mandate final action; a remedy (discretionary or mandatory) is contingent on the ultimate finding of the court.
The same is typically trae of administrative processes. Commonly, an agency seeking to take action adverse to the interests of an affected party brings a charge that leads to a hearing process; only after the hearing does the final agency action result in formal findings and a resultant order.21 The government’s interpretation *169of the statute — and the regulations it has issued that implement it — is essentially consistent with this model. The plaintiffs’ interpretation departs from it, by insisting that a preliminary conclusion sufficient to trigger a full-dress hearing should be treated as an agency “finding” that mandates action.
Fourth, interpreting the statute to mandate action upon a “finding” that is not the result of the required hearing presents the problem of identifying when and how such a finding has been made. Under the plaintiffs’ interpretation, the “finding” that would trigger these mandatory consequences is not, as in the normal understanding of an agency finding, a formal decision embodied in documentary form. The most plausible place to look for a formal finding that precedes and therefore could trigger a hearing under the plaintiffs’ interpretation is in the NOOHs issued by the CVM in 1977 which set forth the scientific conclusions of the Director regarding the safety issues affecting animal antibiotics and initiate the process by which the agency can withdraw approval of their subtherapeutic use.
But if the NOOHs contain or embody the findings on which plaintiffs rely, plaintiffs are confronted with the problem that those NOOHs have been withdrawn. Nothing in the statute or regulations explicitly restricts the FDA’s ability to withdraw an NOOH after it has issued. Thus, at this moment, there is no operative document that contains any formal finding, final or preliminary, that any use of animal antibiotics pose health threats to humans. Accordingly, the plaintiffs must, and do, argue that the withdrawal of the NOOHs does not effectively withdraw the finding that was documented in them. They reason that “findings” need not be reflected in any one document but rather comprise the FDA’s considered collective judgment about the science underlying antibiotic resistance and its effects on human safety. To withdraw the findings, plaintiffs argue, the FDA must publicly recant its earlier position on the safety of the use of antibiotics in animal feed. According to plaintiffs, the agency’s continued insistence, up to and including in briefing and oral argument on this appeal, that such use of antibiotics does pose risks for humans actually “reaffirmed” the findings first announced in the 1977 NOOHs. In short, the “finding” that subtherapeutic uses of antibiotics in animal feed is not shown to be safe resides not in any formal legal conclusion but in the scientific judgment of the relevant FDA officials, current and past, that such uses may be dangerous. By the plaintiffs’ argument, once the Secretary reaches a conclusion that a drug use is not shown to be safe, she is required to act on that opinion.
The withdrawal of the NOOHs, however, simply makes more stark a problem inherent in the plaintiffs’ argument. The underlying logic of the plaintiffs’ position is that the finding of the Secretary that triggers a hearing must precede even the NOOH itself, for it is this finding that triggers the obligation of the FDA to hold a hearing that, assuming that it results in yet another finding adverse to the drug, is the precursor to a mandatory order withholding approval of the drug. In other words, once the Secretary reaches a certain conclusion, an NOOH must issue, and a hearing must commence.
*170That interpretation is problematic. Administrative findings, whether or not preceded by adversarial evidentiary hearings, are ordinarily clothed in the garb of decision, and reflect a formal determination. The fact that plaintiffs argue that the findings originated with the 1977 NOOHs underscores that conclusion, since the NOOHs have the level of formality we typically expect findings to have. But if the NOOHs embody (or contain) the requisite findings, and revocation of the NOOHs does not suffice to withdraw them, where do the findings exist? In the thoughts and beliefs of the Secretary or Commissioner? Scattered across various agency documents reflecting such thoughts?
That is not merely a formal or metaphysical point, but an intensely practical one. By the language of the statute, once a finding is made, agency action is mandatory, and in default of that action, the courts may compel the agency to act. Under the government’s interpretation of the statute, the mandate that the courts are to enforce is straightforward. If, after holding a hearing and reviewing the evidence presented, the agency formally finds that a particular use of an animal drug has not been shown to be safe for humans, but fails to withdraw approval of the use of that drug and instead adopts some other approach for dealing with the prospective danger, the courts must enforce the congressional mandate and require the Secretary to withdraw approval.22
Under the plaintiffs’ interpretation, in contrast, the courts must first determine whether an entirely subjective and unexpressed finding has been made during internal agency deliberations. On the facts of this case, plaintiffs would have us seek such a finding in the now-withdrawn 1977 NOOH, and would have us conclude that the finding continues to exist based on various statements of FDA representatives in public and before the Congress, in litigation, and in the actions taken by the FDA to encourage voluntary reductions in animal antibiotic use. That is itself sufficiently problematic, but in principle plaintiffs’ position would permit lawsuits contending that the Secretary or her delegates have actually made findings that remain entirely unexpressed in any formal document, because they have formed opinions based on internal agency deliberations or on a review of scientific studies.23
Fifth, the traditional model of enforcement action described above contemplates considerable discretion on the part of an agency to decide, for prudential reasons, whether to initiate action or not, and whether to desist from proceeding before a final conclusion is reached. Such discretion is a typical and often necessary feature of the administrative process. Agencies have many responsibilities, and limited resources. Deciding whether and when to deploy those resources in an arduous, contested adversarial process is an important *171and difficult responsibility. It is rare that agencies lack discretion to choose their own enforcement priorities. Indeed, the Supreme Court has long applied a presumption against judicial review of agency decisions declining to proceed with enforcement actions because such decisions are, for purposes of the Administrative Procedure Act (“APA”), “committed to agency discretion.” Heckler v. Chaney, 470 U.S. 821, 882-88, 105 S.Ct. 1649, 84 L.Ed.2d 714 (1985), quoting 5 U.S.C. § 701(a)(2).
Plaintiffs’ interpretation of § 360b(e)(l) would deny that discretion to the FDA. Were the “finding” that requires the withdrawal of approval located not in the final decision of the Commissioner at the conclusion of a hearing at which all relevant evidence is publicly presented, but in a determination by the head of the CVM, based on an internal consideration of studies conducted by the agency or in the academic literature, that the scientific evidence warrants initiating a hearing so that the Commissioner might eventually reach such a final decision, the agency would be required to take irrevocable action whenever the CVM forms such an opinion that a drug is not shown to be safe, regardless of whether the FDA believes that proceeding further is worth the diversion of resources from other agency priorities. Ordinarily, administrative discretion is at its zenith when an agency decides whether to initiate enforcement proceedings. The government’s position is consistent with this longstanding discretion; the plaintiffs’ position is not.
In canvassing these various principles and practices, we do not suggest that they are mandatory and inescapable presumptions about administrative law. Administrative procedure is flexible, and different approaches may be appropriate in different contexts. We have pointed to what we believe are the more common understandings or expectations about agency findings, orderly procedure, administrative discretion, and judicial review. We are confident that numerous exceptions and counter-examples exist. More importantly, it is beyond doubt that Congress has the power to alter these assumptions, in any particular case or in general, by adopting legislation that imposes contrary mandates on administrative agencies.24 Moreover, given the preeminent importance of health and safety in the usage of powerful bioac-tive substances such as human and animal drugs, it would hardly be surprising for Congress to impose limits on traditional agency discretion or to mandate actions protective of human safety. But the issue here is not whether Congress can impose the sort of mandate plaintiffs would find in the statute — of course it can — but whether Congress has done so.
E. Summary
Our survey of the text, the context, the regulations, and the background legal principles leave us firmly persuaded that Congress has not required the FDA to hold hearings whenever FDA officials have scientific concerns about the safety of animal drug usage, that the FDA retains the discretion to institute or terminate proeeed-*172ings to withdraw approval of animal drugs by issuing or withdrawing NOOHs, and that the statutory mandate contained in § 360b(e)(l) applies to limit the FDA’s remedial discretion by requiring withdrawal of approval of animal drugs or particular uses of such drugs only when the FDA has made a final determination, after notice and hearing, that the drug could pose a threat to human health and safety.
That conclusion begins, as it must, with the text of the statute. Although the text is not unambiguously clear, we believe that the FDA put forth the more natural reading. The statute requires the FDA to withdraw approval of an animal drug only “after due notice and opportunity for hearing” has been afforded, and then only “if the Secretary finds” that the drug is not shown to be safe. 21 U.S.C. § 360B(e)(l). That language most naturally refers to a finding that is issued as a result of the hearing. That interpretation, moreover, avoids injecting a second, unexpressed “finding” into the sequence of events mentioned in the statute.
Although the regulations implementing the statute do not directly address the question of interpretation posed by the plaintiffs, and contain at least some language that arguably supports the plaintiffs’ reading of some terms in the statute, the overall thrust of the regulations is consistent with the government’s interpretation, and with what we regard as the more natural reading of the statutory language. Moreover, the procedure set forth in those regulations, and our reading of the text, are consistent with common assumptions about agency procedure. Under that procedure, relevant experts within the agency (the staff of the CVM) first assess the scientific issues regarding the risks and benefits of the drug, and a high-ranking agency official (the Director of the CVM) exercises discretion to institute a proceeding that can lead to the revocation of approval of the drug. Then (if the sponsor of the drug requests a hearing and raises genuine issues of material fact about the preliminary conclusions set forth in the NOOH) the staff proceeds to present evidence at a hearing featuring the separation of functions between the “prosecuting” officials instituting the hearing and the objective decisionmaker who will hear the evidence. At the conclusion of that hearing, that decisionmaker issues findings that must be approved by a higher-ranking official, the Commissioner of the FDA. If the ultimate agency finding, which is subject to judicial review, is that the drug is not shown to be safe, the statute permits only one remedy — withdrawal of approval.25
Our interpretation of the statute is consistent with the regulations and with conventional procedure. That alone does not make it correct; Congress undoubtedly has the power to alter those conventions. We believe, however, that if Congress intended to do so, and to mandate the commencement of the notice and hearing process whenever the agency staff formed a scientific opinion adverse to the drug, it would have stated those intentions explicitly. Far from doing so, it has utilized language that is not only consistent with the traditional administrative process, but that is more naturally read as adopting it.
II. The Citizen Petitions
Alternatively, plaintiffs argue that even if the FDA is not required to proceed with *173hearings, its decision denying plaintiffs’ 1999 and 2005 citizen petitions and withdrawing the 1977 NOOHs represented arbitrary or capricious agency action in violation of the APA. See 5 U.S.C. § 706(2)(A). In particular, plaintiffs argue that the FDA’s denials were based on factors not explicitly mentioned by the statute, namely cost, time, and a preference for voluntary compliance over adversary proceedings.
The FDA’s notice withdrawing the 1977 NOOHs sets out the reasons for the action:
FDA is taking this action, and closing the corresponding dockets, because: FDA is engaging in other ongoing regulatory strategies developed since the publication of the 1977 NOOHs with respect to addressing microbial food safety issues; FDA would update the NOOHs to reflect current data, information, and policies if, in the future, it decides to move forward with withdrawal of the approved uses of the new animal drugs described in the NOOHs; and FDA would need to prioritize any withdrawal proceedings (for example, take into account which withdrawal(s) would likely have the most significant impact on the public health) if, in the future, it decides to seek withdrawal of the approved uses of any new animal drug or class of drugs.
Withdrawal of Notices of Opportunity for a Hearing: Penicillin and Tetracycline Used in Animal Feed, 76 Fed.Reg. 79697, 79697 (Dec. 22, 2011). The letters in which the FDA finally denied of the citizen petitions further elaborate on the decision to withdraw the NOOHS and deny the petitions, stating “reviewing safety information for antimicrobial drugs approved before 2003, and pursuing withdrawal proceedings in some cases, would take many years and would impose significant resource demands on the [FDA].” J.A. at 622; J.A. at 627. In the letters the FDA describes its plan to “work[] cooperatively” with the animal pharmaceutical industry to “ensure] the judicious use of medically important antimicrobial drugs in food-producing animals.” Id.
In arguing that such denial is arbitrary or capricious, plaintiffs claim that the FDA ignored the reams of scientific data presented in the petitions and that the reasons given by the FDA are illegitimate because they are orthogonal to what plaintiffs persuaded the district court is the governing criterion described in the statute: “whether the drugs at issue pose a threat to human health.” See NRDC II, 872 F.Supp.2d at 338.
Plaintiffs argue that this case is best analogized to Massachusetts v. EPA, 549 U.S. 497, 127 S.Ct. 1438, 167 L.Ed.2d 248 (2007), in which the Supreme Court invalidated the denial of a petition seeking to require the Environmental Protection Agency (“EPA”) to regulate greenhouse gases. But Massachusetts dealt with a much different statutory provision, one which unambiguously compelled agency action. In that case, a group of states, municipalities, and non-profit organizations sought to force the EPA to regulate four greenhouse gases as air pollutants under Section 202(a)(1) of the Clear Air Act, which provides that:
The [EPA] Administrator shall by regulation prescribe ... standards applicable to the emission of any air pollutant from any class or classes of new motor vehicles or new motor vehicle engines, which in his judgment cause, or contribute to, air pollution which may reasonably be anticipated to endanger public health or welfare.
42 U.S.C. § 7521(a)(1) (emphasis added). The Supreme Court agreed with the plaintiffs that the text of the Clean Air Act required the EPA to regulate greenhouse *174gases. In reaching that conclusion, the Court understood the “judgment” contemplated by the statute as limited to the scientific question whether a particular pollutant contributed to air pollution. Put differently, the Supreme Court read the Clean Air Act not to grant the EPA discretion to choose to regulate only those pollutants that it deemed feasible or wise to regulate. As the Court had it, “the use of the word ‘judgment’ is not a roving license to ignore the statutory text.” Massachusetts, 549 U.S. at 533, 127 S.Ct. 1438. Once the EPA determined that carbon dioxide contributed to air pollution, the Court concluded, the statute required it to regulate the emission of that gas.
Massachusetts v. EPA is therefore fully distinguishable from the present case. The Clean Air Act limited the EPA Administrator’s “judgment” to the scientific question of whether the pollutant in question causes dangerous air pollution; nothing in § 360b(e)(l) limits the considerations that the FDA may take into account in deciding whether to initiate the hearing process by issuing an NOOH. Moreover, unlike the Clean Air Act, which explicitly and unambiguously requires the regulation of pollutants (“The Administrator shall by regulation prescribe ... standards”), as explained above, § 360b(e)(l) does not mandate that the FDA take any action until and unless certain findings are made after a hearing.26 In short, the Clean Air Act provision at issue in Massachusetts v. EPA unambiguously required the EPA to undertake action to create emission standards (leaving it to the EPA’s expertise to determine the substance of the standards) whenever it forms a scientific judgment that a particular pollutant contributes to dangerous air pollution, while the provision of the FDCA at issue in this case requires the FDA to take a specific remedial step when, after a hearing, it has made certain findings, without imposing any absolute requirement that the agency investigate the need for withdrawing approval of animal drugs under any particular circumstance.
The present case is therefore more analogous to New York Public Interest Research Group v. Whitman, 321 F.3d 316 (2d Cir.2003), in which we interpreted section 502(i) of the Clean Air Act, 42 U.S.C. § 7661a(i)(l). That provision required the Administrator of the EPA to give notice and, if appropriate, impose the relevant sanctions, “[wjhenever the Administrator makes a determination that a permitting authority is not adequately administering and enforcing a program ... in accordance with the requirements of this sub-chapter.” Id. (emphasis added) We held that the use of the word “determination” “grants discretion.” Whitman, 321 F.3d at 330. Rejecting the view that the EPA was required to issue a notice of deficiency whenever it found defects in a state permitting program, we noted that “Congress could have fashioned a regime under which, for example, an interested party could initiate the process leading to a determination of whether ‘a permitting authority is adequately administering and enforcing a program,’ ” but that by refer*175ring to a “determination” on the part of the agency, Congress left it to the discretion of the EPA Administrator whether and when to initiate enforcement proceedings. Id. at 331, quoting 42 U.S.C. 7661a(i)(l).27
For the reasons set forth above, we conclude that the decision whether to institute or terminate a hearing process that may lead to a finding requiring withdrawal of approval for an animal drug is a discretionary determination left to the prudent choice of the FDA.28
On that basis, it is relatively easy for us to accept the FDA’s determination that its preferred program of voluntary compliance offers greater prospect for immediate and significant reductions in animal antibiotic use than the pursuit of a potentially contentious withdrawal hearing. That is the sort of prudential judgment better suited to expert administrators than to federal judges. We are bolstered in this conclusion by the nature of the problem confronted by the FDA. Nothing in the NOOHs suggests that penicillin and tetracycline, when administered to animals, are inherently dangerous to human health; such antibiotics are widely, effectively, and beneficially used in human medical care. And while we are hesitant, for the reasons set forth above, to ascribe scientific conclusions to the FDA based on our reading of a melange of different studies, regulatory documents, and litigation positions, it appears clear that while the agency regards the indiscriminate and extensive use of such drugs in animal feed as threatening, it does not necessarily believe that the administration of antibiotics to animals in their feed is inherently dangerous to human health.
Under these circumstances, we cannot conclude that it is arbitrary or capricious for the FDA to pursue polices intended to reduce the use of animal feed containing antibiotics through a variety of steps short of withdrawing approval for the use of antibiotics in feed via a protracted administrative process and likely litigation. As it was neither arbitrary nor capricious for the FDA to deny the petitions for the reasons it did, the district court’s decision to the contrary was error.
In letters recently submitted to the Court by the government, the government notes that the FDA is “encouraged” by the “overwhelmingly cooperative” reaction of the animal feed industry to the guidelines for voluntary compliance that the agency has issued in lieu of proceeding with the process initiated in 1977 with the issuance of the Penicillin and Tetracycline NOOHs, Gov’t Letter dated March 27, 2014, and asserts that the high level of cooperation by drug manufacturers “demonstrate^] that the cooperative approach ... has *176been effective in enabling FDA to achieve its goals of phasing out the use of medically important antimicrobial drugs for food-production purposes,” Gov’t Letter Dated July 1, 2014. In light of the discussion above, it should be obvious that we express no opinion on the effectiveness of the FDA’s approach to what it agrees is a significant regulatory concern about the overuse of antibiotics in animal feed, and that in determining the issues in this case, we place no weight on the agency’s informal assurances that its program is successful. It is not for us to determine whether the agency has been prudent or imprudent, wise or foolish, effective or ineffective in its approach to this problem. Whether the agency’s long inaction in the face of the dangers highlighted in the 1977 NOOH’s represented politically-inspired foot-dragging or wise caution in developing a cost-effective approach, it was for the agency, and not the courts, to determine how best to proceed.
CONCLUSION
For the foregoing reasons, the decisions of the district court are REVERSED, and the case is remanded to the district court with instructions to deny the plaintiffs’ motion for summary judgment, grant the defendants’ motion for summary judgment, and dismiss the action.
. For another case arising from an effort to force the FDA to limit nontherapeutic uses of antibacterial agents in a different context, see Natural Res. Def. Council v. FDA, 710 F.3d 71 (2d Cir.2013).
. The term “new animal drug” is defined in 21 U.S.C. § 321(v); see note 11, infra, for the text of that section.
. Generic-drug applications receive a slightly different label, "abbreviated NADA.”
. "Subtherapeutic” uses are those that seek "increased rate of [weight] gain, disease prevention!,] etc.,” as opposed to uses to treat illnesses or other pathological conditions. 21 C.F.R. § 558.15(a). Other sources prefer the term "nontherapeutic,” for the same meaning.
. The Bureau of Veterinary Medicine is now known as the Center for Veterinary Medicine. In the interest of simplicity, we use the abbreviation CVM to refer to that subdivision of the FDA regardless of its official title at any given time.
. The plaintiffs are the Natural Resources Defense Council, Inc., the Center for Science in the Public Interest, the Food Animal Concerns Trust, Public Citizen, Inc., and the Union of Concerned Scientists, Inc.
. Plaintiffs also pled claims pursuant to § 706(1) of the Administrative Procedure Act ("APA”), which generally authorizes courts presented with challenges to agency inaction to "compel agency action unlawfully withheld or unreasonably delayed.” 5 U.S.C. § 706(1). Because the essence of plaintiffs’ APA claim is that the FDCA requires the FDA to hold the hearings described in the 1977 NOOHs, the question posed by § 706 is identical to that posed by the text of § 360b(e) itself.
.The parties did not conduct discovery.
. The parties had consented to trial before a
. "New animal drug” is elsewhere defined, subject to limited exceptions, as "any drug intended for use for animals other than man, including any drug intended for use in animal feed but not including such animal feed, ... the composition of which is such that such drug is not generally recognized, among experts qualified by scientific training and experience to evaluate the safety and effectiveness of animal drugs, as safe and effective for use under the conditions prescribed, recommended, or suggested in the labeling thereof.” 21 U.S.C. § 321(v) (2008).
. Although the statute refers to the Secretary of Health and Human Services, the Secretary has delegated her duties under the FDCA to the Commissioner of the Food and Drug Administration.
. The hearing would have an Alice-in-Wonderland quality (sentence first, trial afterward) unless it were understood that the Secretary may only ultimately withdraw approval if the evidence presented at the hearing warrants a finding that the drug is not shown to be safe. There would be no purpose to a hearing if the agency’s initial internal conclusion of itself mandated withdrawal, regardless of the outcome of the hearing.
. In his thoughtful dissenting opinion, Judge Katzmann correctly notes that "[W]e begin ... any exercise of statutory construction with the text of the provision in question, and move, as need be, to the structure and purpose of the Act in which it occurs.’’ (Dissenting opinion, post, at 177, quoting N.Y. State Conference of Blue Cross & Blue Shield Plans v. Travelers Ins. Co., 514 U.S. 645, 655, 115 S.Ct. 1671, 131 L.Ed.2d 695 (1995)). This principle guides our analysis, and confirms our respectful disagreement with Judge Katz-mann’s conclusion regarding the interpretation of the statute.
Insofar as the dissent’s analysis refers to the purpose of the statute, it hinges on the unquestioned goal of Congress to protect human health. But this is largely a red herring. As Judge Katzmann acknowledges (Dissenting opinion, post, at 178), that goal does not require us to interpret any ambiguity in the statute in the manner that we think is most conducive to protecting the public health; the statute reflects in its language particular judgments about how that goal should be pursued and when it must yield to or be balanced with other concerns. Our views regarding how the FDA can best serve its mission of protecting human health through the use of drugs to treat animals or the analytic import of other prospectively ambiguous Congressional statutes must defer to the most reasonable reading of the text before us. Section 360b(e)(l) is most "naturally read” (Dissenting opinion, post, at 183) in the manner that makes best sense of the statutory language itself. That in turn compels us to conclude that where a statute explicitly considers only a single finding and directs that any such finding mandates the agency to take decisive action, the statute only involves a single finding by that agency.
Insofar as the dissent addresses the statutory structure, it primarily relies on a perceived parallel between the procedures for initial approval of a drug and withdrawal of that approval. But this attempted parallel ignores the fact that a withdrawal procedure occurs after a drug has already been found safe-&emdash;a difference that amply justifies a different process.
. Put another way, plaintiffs’ interpretation is internally inconsistent. On the one hand, it relies on the literal language of the statute to insist that upon the "finding” of lack of showing of safety — by which plaintiffs mean the preliminary internal conclusion of the agency that leads to the issuance of an NOOH— withdrawal of approval is mandatory, but on the other they acknowledge that that preliminary “finding” does not and cannot in fact mandate withdrawal of approval, because it leads only to a hearing that may or may not result in a finding that the drug is not shown to be safe. Plaintiffs insist that the "finding” or preliminary assessment of the agency mandates that the agency hold a hearing, but the finding referred to in § 360b(e)(l) requires that the Secretary issue an order withdrawing approval of the drug. Only a finding that is made after notice and an opportunity for a hearing can have this effect.
. 21 U.S.C. § 360b(d)(l) states: "If the Secretary finds, after due notice to the applicant in accordance with subsection (c) of this section and giving him an opportunity for a hearing, in accordance with said subsection, that [one of nine specified conditions is satisfied,] he shall issue an order refusing to approve the application. If, after such notice and opportunity for hearing, the Secretary finds that [those conditions are not met], he shall issue an order approving the application.”
. Specifically, the Secretary is authorized to approve an application for a new animal drug if she “finds that none of the grounds for denying approval specified in [§ 360b(d) ] applies.” 21 U.S.C. § 360b(c)(l).
. 21 C.F.R. § 5.84 has been formally withdrawn in an amendment of the regulations governing the FDA organizational structure, promulgated by a rule that is "editorial in nature.” Revision of Organization and Conforming Changes to Regulations, 77 Fed.Reg. 15961-02 (Mar. 19, 2012) (codified at 21 C.F.R. Chapter I). The substance of the delegation from the Secretary to the CVM contained within 21 C.F.R. § 5.84 is now authorized by FDA Staff Manual Guides, Volume II — Delegation of Authority, SMG § 1410.503 (Feb. 24, 2011). As the modification to the regulations was editorial rather than substantive, the language of the delegation remains the same, and both parties implicitly agree that § 5.84 remains in effect, we address the regulation here.
. As the district court noted, the Commissioner of the FDA ratified the judgment reflected in the 1977 NOOHs by concurring — in a policy statement published in the Federal Register — with the Director's 1983 decision not to rescind those NOOHs. Penicillin and Tetracycline (Chlortetracycline and Oxytetra-cycline) in Animal Feeds; Denial of Petitions, 48 Fed.Reg. 4554, 4556 (Feb. 1, 1983) ("The Commissioner has reviewed the Director's decision and concurs with it. In addition, for the reasons discussed above, the Commissioner has decided that he will not withdraw the advance notice of hearing or the proposal to restrict therapeutic approvals to prescription use, but will hold them in abeyance.”).
. Looking beyond the specific context of the Federal Food, Drug, and Cosmetic Act, Judge Katzmann cites a number of statutes in which Congress has used language similar to that at issue here. (Dissenting opinion, post, at 183). Significantly, however, he cites no judicial interpretation of this language that supports plaintiffs' readings (or, for that matter, that supports ours). The fact that Congress has created similar ambiguities in other contexts tells us nothing about how to resolve the ambiguity. Similarly, we are unpersuaded by the dissent’s suggestion that the Supreme *168Court offhandedly endorsed plaintiffs' view of § 360(b)(1)(B) in Brown & Williamson (Dissenting opinion, post, at 180-81). In the cited passage, 529 U.S. at 134, 120 S.Ct. 1291, the Court simply repeats, in a slightly condensed form, the ambiguous language of the statute itself.
. To take only one example of this familiar process, when the Securities and Exchange Commission seeks to impose a cease-and-de*169sist order against a corporation for violating Section 21C of the Securities Exchange Act of 1934, the Commission’s staff issues a notice of a hearing prior to the issuance of such an order. It is the final decision of the Commission, not the charges contained in the notice, that constitute the “findings” of the agency, lead to the issuance of the order, and trigger judicial review.
. Moreover, upon the issuance of findings for or against the demonstration of the safety of the drug at the conclusion óf a hearing, judicial review of whether the findings are adequately supported by the record will be available.
. The dissent disavows any reliance on the subjective views of FDA officials (Dissenting opinion, post, at 185), but then falls back on essentially the same argument about what the agency really believes by arguing that the now-withdrawn NOOHs were merely the “medium” for the "message” they contained: the prior internal agency conclusions that penicillin and tetracycline had not after all been shown to be safe (Dissenting opinion, post, at 185-86). The dissent concludes that withdrawal of the formal document on which it relies as a basis for judicial review is insufficient unless the agency has actually changed its mind, and deduces from a variety of sources that it has not.
. Such was the situation in the recently decided Cook v. FDA, 733 F.3d 1 (D.C.Cir.2013). There, the D.C. Circuit found the FDA failed to enforce a statute that included a clear Congressional mandate for the FDA to take action in particular circumstances. The FDA argued that because enforcement is a matter of agency discretion, the decision to not enforce the statute was not subject to judicial review. Id. at 5. The court found the relevant statute to be "unambiguously binding” on the FDA in mandating that particular actions be taken. Id. at 7 (citation omitted). Thus, the case addressed an instance of explicit legislative instruction stripping an agency of discretion. Such clear legislative instruction is noticeably absent here.
. Our reading thus emphatically recognizes the mandatory language of the statute. Where the FDA would otherwise have considerable discretion to take whatever action might be appropriate to protect the public safety in light of the results of the hearing, Congress has specifically mandated in § 360b(e)(l) that only one response is appropriate.
. Plaintiffs' reliance on Brown & Williamson is also misplaced. In that case, the Supreme Court held that the FDA lacked statutory authority to regulate the sale of tobacco products. 529 U.S. at 125-26, 120 S.Ct. 1291. In the course of its analysis, the Court noted that if it had statutory jurisdiction to regulate tobacco products, FDA would have been required by its organic statute to remove them from the market altogether. Id. at 135-36, 120 S.Ct. 1291. There was no question in that case, however, that FDA had made ''findings” about the safety of tobacco before issuing a final rule governing youth access to tobacco products. The Supreme Court did not address whether, if tobacco fell within its jurisdiction, the FDA would have been required to initiate, or forbidden from abandoning, a course of hearings that might or might not have resulted in such findings.
. As further evidence of the absence of a statutory mandate, we noted the lack of any time statutory time limits on the agency’s action, 321 F.3d at 331 n. 8, a factor also applicable here.
. We respectfully disagree with Judge Katz-mann's assertion (Dissenting opinion, post, at 187-88) that because we find Whitman to offer a helpful analogy to determine whether the FDA abused its discretion, we are implicitly equating a withdrawal action to an enforcement action. Whitman provides guidance in that the relevant statutory language in that case ("Whenever the Administrator makes a determination,” 321 F.3d at 330, quoting 42 U.S.C. § 7661(a)(i)(l)) left the mandatory agency action conditional upon a discretionary agency finding, much like the statutory provision at issue here (action is required only "if the Secretary finds” that certain conditions adhere). The relevant parallel in these cases is one of analogical statutory construction that leaves action dependent upon agency discretion, a construction distinguishable from the unequivocal imperative in the statute at issue in Massachusetts v. EPA. Whether a withdrawal action is an enforcement action is not relevant to our conclusion.
. To be more precise, § 360b(a) declares that a new animal drug "shall, with respect to any particular use or intended use of such drug, be deemed unsafe” unless it has been approved by the FDA or a special exception applies. Section 351(a)(5) then declares that any new animal drug deemed unsafe under § 360b(a) shall be deemed "adulterated,” and § 331(a) prohibits introducing any adulterated drug into interstate commerce. Throughout the discussion below, I omit similarly tangential details of the FDCA’s intricate statutory scheme.