Partial Concurrence by Judge WATSON
OPINION
IKUTA, Circuit Judge:
Glacier Fish Co. challenges the fee imposed on it by the National Marine Fisheries Service (NMFS) pursuant to the agency’s cost recovery program developed under the Magnuson-Stevens Fishery Conservation and Management Act (Mag-nuson-Stevens Act, or MSA). We conclude that the agency had the authority to require Glacier to pay a cost recovery fee but that the .agency’s calculation of the amount of the 2014 cost recovery fee was inconsistent with its own regulations. We therefore reverse the district court’s grant of summary judgment in part and remand to the agency.
I
The Magnuson-Stevens Act created eight Regional Fishery Management Councils.1 Each council creates a fishery [1116]*1116management plan, which must contain conservation and management measures and an assessment of the maximum sustainable yield from each fishery. 16 U.S.C. § 1853(a). Beginning in 1990, the councils were given the discretion to use “a limited access system for the fishery in order to achieve optimum yield.” Id. § 1853(b)(6). A limited access system allows only those entities that satisfy certain eligibility criteria or requirements contained in a fishery management plan to participate in a fishery. Id. § 1802(27).
In 2007, Congress reauthorized the Magnuson-Stevens Act with amendments that, among other things, were intended to encourage market-based fishery management through “limited access privilege programs” (referred to by NMFS as a LAPP). The term “limited access privilege” is defined as “a Federal permit, issued as part of a limited access system under section 1853a of this title to harvest a quantity of fish ... representing a portion of the total allowable catch of the fishery that may be received or held for exclusive use by a person.” 16 U.S.C. § 1802(26). Thus, a limited access privilege program (which must be part of a limited access system) is a program in which fishery participants obtain a Federal permit “to harvest a certain portion of the total catch allowed for a particular species.” Pac. Coast Fed’n of Fishermen’s Ass’ns v. Blank, 693 F.3d 1084, 1088 (9th Cir. 2012). One way to implement such a program is to distribute the allocated portion through “individual fishing quota” (IFQ), or quota shares (i.e., a specified percentage of the total catch). 16 U.S.C. §1802(23). But councils need not adopt IFQ programs; they retain discretion to implement limited access privilege programs in different ways. See id. §§ 1853a, 1802(26).
Any council that elects to implement a limited access privilege program must also implement a cost-recovery program. Id. § 1853a(e). The Magnuson-Stevens Act requires councils to “develop a methodology and the means to identify and assess the management, data collection and analysis, and enforcement programs that are directly related to and in support of the program,” id. § 1853a(e)(l), and then provide “for a program of fees paid by limited access privilege holders that will cover the costs of management, data collection and analysis, and enforcement activities,” id. § 1853a(e)(2). The Secretary of Commerce is authorized to “collect a fee to recover the actual costs directly related to the management, data collection, and enforcement of any limited access privilege program.” Id. § 1854(d)(2)(A)(i). The fee “shall not exceed 3 percent of the ex-vessel value2 of fish harvested under any such program, and shall be collected at either the time of the landing, filing of a landing report, or sale of such fish during a fishing season or in the last quarter of the calendar year in which the fish is harvested.” Id. § 1854(d)(2)(B).
Once a regional council has prepared a fishery management plan for each fishery within its jurisdiction that requires such a plan, it submits the plan and any proposed regulations to the Secretary of Commerce. Id. § 1852(h)(1). The Secretary must review the plan to determine whether it is consistent with statutory requirements, id. § 1854(a)(1)(A), and publish a notice of proposed rulemaking in the Federal Register, id. § 1854(a)(1)(B). This publication [1117]*1117starts a public notice and comment period. Id. If the Secretary approves the plan, the Secretary must review the council’s proposed regulations for consistency with its fishery management plan and other law. 16 U.S.C. § 1854(b). The Secretary must publish these regulations as well for public comment. Id. The Secretary has delegated her responsibilities under the Act to NMFS, which is housed in the National Oceanic and Atmospheric Administration in the Department of Commerce.
One of the eight regional councils established by the Magnuson-Stevens Act is the Pacific Fishery Management Council (Pacific Council), which consists of representatives from California, Oregon, Washington, and Idaho and covers the fisheries seaward of those states. 16 U.S.C. § 1852(a)(1)(F). One of those fisheries is the Pacific groundfish fishery, which “extends 200 miles into the Pacific Ocean, along the coasts of California, Oregon, and Washington, and includes more than 90 species of fish that dwell near the sea floor.” Pac. Coast Fed’n of Fishermen’s Ass’ns, 698 F.3d at 1088. The Pacific groundfish fishery is comprised of three sectors: the catcher-processor (C/P) sector, which consists of trawl fishing vessels that catch and process whiting on board; the shoreside sector, which consists of vessels that catch whiting and deliver the catch to processors on land; and the mothership sector, which consists of vessels that catch whiting and deliver the catch to processors at sea. The Pacific Council first developed the Pacific Coast Groundfish Fishery Management Plan (Groundfish Management Plan) in 1982. The plan covers the Pacific whiting, the species of fish at issue here, among other groundfish.
In 1994, Amendment 6 to the Groundfish Management Plan limited participation in the Pacific groundfish catcher-processor sector by requiring each vessel to obtain one ■ of a small number of limited entry permits.3 The Council also provided certain allocations of whiting to each sector of the fishery; for instance, in 1997, the Council allocated 34 percent of the allowable catch to.the catcher-processor sector, 42 percent to the shoreside sector, and 24 percent to the mothership sector. Whiting Allocation Among Nontribal Sectors, 62 Fed. Reg. 27519-01, 27520 (May 20, 1997). In addition, the Council established a short season for whiting and allowed permitted vessels to harvest whiting only frota the time the season opened until the time the catch limit was reached. This management structure led to a so-called “race for fish.” A vessel with a limited entry permit had an incentive to make an intense, concentrated effort at the beginning oí the season to harvest the largest possible portion of the catcher-processor allocation of whiting before the window closed. ■
In an attempt to eliminate this “race for fish,” catcher-processors formed a private cooperative, the Pacific Whiting Conservation Cooperative (PWCC), in 1997. Through private agreements, each member of the PWCC agreed to limit its own whiting harvest to a certain percentage of the catcher-processor secfor whiting allocation. By agreeing to apportion shares of the whiting allocation in advance, the members no longer needed to race to catch their portion of the catcher-processor sector’s allocation. This private solution was effective, and NMFS acknowledged that the catcher-processor sector no longer raised the “race for fish” problem.
But because other sectors had not successfully ended the race for fish, in 2003, the Pacific Council and NMFS began developing a limited acceás privilege program for the groundfish fishery as a whole [1118]*1118(including the catcher-processor sector). The program’s primary goal was to implement an IFQ program (or quota share system) for the shoreside sector. In January 2004, NMFS published a notice of proposed rulemaking, which stated that the Pacific Council was considering implementing a limited access privilege program in the form of a “trawl rationalization program” for the Pacific groundfish fishery. 69 Fed. Reg. 1563-01 (Jan. 9, 2004). The Pacific Council then engaged in a lengthy process to develop an amendment to the Groundfish Management Plan that would implement the trawl rationalization program. Finally, in 2009, the Pacific Council submitted its proposal for a trawl rationalization program to the Secretary as Amendment 20 to the Groundfish Management Plan. NMFS published the proposed amendment for comment in May 2010, see Amendments 20 and 21, Trawl Rationalization Program, 75 Fed. Reg. 26,702-01 (May 12, 2010), and the proposed regulations for implementing Amendment 20 in June 2010, see Amendments 20 and 21, Trawl Rationalization Program, 75 Fed. Reg. 32,994-01 (June 10, 2010). After a round of notice and comment, NMFS adopted Amendment 20 to the Groundfish Management Plan and promulgated implementing regulations. See Amendments 20 and 21, Trawl Rationalization Program, 75 Fed. Reg. 60,-868-01 (Oct. 1, 2010).
The final trawl rationalization program (Amendment 20) established an IFQ program for the shoreside sector. An IFQ program is “a quota system where each quota share could be harvested at any time during an open season.” Advance Notice of Proposed Rulemaking Regarding a Trawl Individual Quota Program and to Establish a Control Date, 69 Fed. Reg. 1563-01, 1563 (Jan. 9, 2004). “Participants in the fishery (i.e., those who already had a limited entry permit allowing them to fish) would need to obtain a quota share permit as well in order to receive a share of the allowable catch.” Pacific Dawn, No. 14-15224, draft at 8 (citing 16 U.S.C. §§ 1853a, 1802(26)).
But because the catcher-processor sector had been operating successfully as a cooperative, Amendment 20 adopted a different approach for the catcher-processor sector. Rather than requiring each catcher-processor to obtain an IFQ or quota share permit in order to receive its share of the catch, Amendment 20 allowed the members of the coop (i.e., all the participants in the catcher-processor sector) to obtain a single “coop permit” in order to receive the catcher-processor sector’s entire allocation of catch. The members of the coop would then continue to distribute that allocation amongst themselves through their private agreements. Amendment 20 termed this a “coop program.” The participants in the catcher-processor sector would be required to obtain individual IFQs or quota share permits only if the coop dissolved. 50 C.F.R. § 660.160(h)(4)(ii).
The final regulations defined the “coop program” as “a limited access program that applies to vessels in the C/P sector of the Pacific whiting at-sea trawl fishery” that had formed a voluntary coop. Id. § 660.160(a). A “catcher/processor coop” is defined as “a harvester group that includes all eligible catcher/processor at-sea Pacific whiting endorsed permit owners who voluntarily form a coop and who manage the catcher/processor-specified allocations through private agreements and contracts.” Id. § 660.111. Under the program, a coop must obtain a “coop permit,” renewable annually. A coop permit “conveys a conditional privilege to an eligible coop entity to receive and manage a coop’s allocation of designated species and species groups.” Id. § 660.25(e)(2). According to NMFS, the coop permit “formally register[s] the coop and its associated members [1119]*1119to harvest and process in the sector.” 75 Fed. Reg. 53380-01, 53392. The permit allows members to engage in fishing of the catcher-processor sector’s allocation during the season. 50 C.F.R. § 660.160(d)(l)(iv). The coop program also required the coop to submit an annual report to NMFS, while the members of the coop are required to submit economic data. Id. §§ 660.160(b)(l)(ii), (b)(2)(h). In sum, Amendment 20 made only a few changes to the catcher-processor sector; it continued to operate as a coop through which members divided their allocation according to their private agreements.
After implementing Amendment 20 and the trawl rationalization program, the Pacific Council began to develop a cost recovery program as required by 16 U.S.C. § 1853a(e)(l). It submitted a proposed program to NMFS in December 2012. On February 1, 2013, NMFS proposed regulations establishing a cost recovery program and solicited public comments. See Cost Recovery, 78 Fed. Reg. 7371-01 (Feb. 1, 2013). On December 11, 2013, NMFS published its final rule, which established the requirements for cost recovery from each of the sectors. See Cost Recovery, 78 Fed. Reg. 75,268-01 (Dec. 11, 2013); 50 C.F.R. § 660.160(e)(5).
The final regulations required members of a catcher-processor coop to pay a percentage of the revenue earned by each vessel (i.e., its ex-vessel value) as a fee to NMFS. NMFS calculated the fee percentage based on DPC, or direct program costs, defined as “the actual incremental costs for the previous fiscal year directly related to the management, data collection, and enforcement” of the catcher-processor sector, 50 C.F.R. § 660.115(b)(l)(I), divided by V, “the total ex-vessel value, as defined at § 660.111, from the previous calendar year attributable to that sector of the trawl rationalization program,” id. § 660.115(b)(1)(h).4
Comment 5 to the proposed regulations asked NMFS to “provide the legal basis for defining the C/P Coop Program as a LAPP [limited access privilege program].” 78 Fed. Reg. 75,268-01, 75,272. NMFS responded that it “decided that the C/P Coop Program was a LAPP during implementation of Amendment 20.” Id. It further explained that “[c]onsistent with the definition of a ‘limited access privilege’ in the MSA (16 U.S.C. § 1802 (26)), the C/P Coop Program is a LAPP under the MSA (16 U.S.C. § 1853a) because it requires a Federal permit for exclusive usd by the coop to harvest a portion of the total allowable catch.” Id. NMFS also acknowledged “that generally the C/P Coop Program management costs are less than those of the other sectors.” Id.
In 2013, the PWCC received a catcher-processor coop permit that stated: “This permit authorizes the entity named above [PWCC] to harvest 100% of the Pacific Whiting and non-whiting allocated to the C/P sector,” which was 34 percent of the total allowable catch. The three coop members, American Seafoods, Trident Sea-foods, and Glacier Fish, then apportioned the allocation among themselves according to their agreements.
To calculate the 2014 fee percentage for purposes of its cost recovery program, NMFS determined its 2013 direct program costs for the catcher-processor sector to be $176,460. NMFS divided the 2013 direct program costs by the total catcher-processor sector ex-vessel value of $16.76 million, which resulted in a fee percentage of 1.1 percent. Thus, each participant in the catcher-processor coop program was re[1120]*1120quired to pay a fee of 1.1 percent of its 2014 revenue.5
On January 9, 2014, Glacier filed this action in district court. The parties filed cross-motions for summary judgment. The district court rejected each of Glacier’s arguments and granted summary judgment to NMFS. Glacier timely appealed.
On appeal, Glacier makes three arguments. First, Glacier argues that because NMFS may collect a fee only from “limited access privilege holders,” 16 U.S.C. § 1853a(e)(2), it is not authorized to collect a cost recovery fee from Glacier for two reasons: (1) a “coop permit” is not a limited access privilege because it does not authorize PWCC to harvest a quantity of fish and it lacks one of the statutory characteristics of a limited access privilege; and (2)it is not the holder of the coop permit. Second, Glacier argues that NMFS failed to implement the cost accounting methodology recommended by the Council in violation of 16 U.S.C. § 1853a(e). Third, Glacier argues that NMFS’s calculation of the 2014 fee was inconsistent with NMFS’s own regulations. We address each of Glacier’s arguments in turn.6
II
The district court had jurisdiction under 28 U.S.C. § 1331, and we have jurisdiction under 28 U.S.C. § 1291. We review the district court’s grant of summary judgment de novo. Fishermen’s Finest, Inc. v. Locke, 593 F.3d 886, 894 (9th Cir. 2010).
We review NMFS’s interpretation of the Magnuson-Stevens Act under the two-step framework of Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842-43, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). See Pacific Coast Fed’n of Fishermen’s Ass’ns, 693 F.3d at 1091; Or. Trotters Ass’n v. Gutierrez, 452 F.3d 1104, 1119 (9th Cir. 2006).
At step one, we consider whether “the intent of Congress is clear.” Chevron, U.S.A., Inc., 467 U.S. at 842, 104 S.Ct. 2778. If “Congress has directly spoken to the precise question at issue,” we “must give effect to the unambiguously expressed intent of Congress.” Id. at 842-43, 104 S.Ct. 2778. We will uphold the agency’s interpretation if it is consistent with that congressional intent. See Or. Trotters Ass’n, 452 F.3d at 1119.
If the statute is “silent or ambiguous” on the precise issue, or susceptible to multiple interpretations, we consider whether Congress delegated authority to the agency “to speak with the force of law when it addresses ambiguity in the statute or fills a space in the enacted law.” United States v. Mead Corp., 533 U.S. 218, 229, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001); see also Nat’l Cable &Telecomms. Ass’n v. Brand X Internet Servs., 545 U.S. 967, 980, 125 S.Ct. 2688, 162 L.Ed.2d 820 (2005) [1121]*1121(“[A]mbiguities in statutes within an agency’s jurisdiction to administer are delegations of authority to the agency to fill the statutory gap in reasonable fashion.”). Congress has delegated such authority “when it provides for a relatively formal administrative procedure,” like notice-and-comment rulemaking. Mead, 533 U.S. at 230, 121 S.Ct. 2164. And where the agency interprets the statute through an exercise of that authority, we must “accept the agency’s construction of the statute” so long as “the implementing agency’s construction is reasonable, ... even if the agency’s reading differs from what the court believes is the best statutory interpretation.” Brand X, 545 U.S. at 980, 125 S.Ct. 2688. An interpretation is reasonable so long as it “reflects a plausible construction of the plain language of the statute and does not otherwise conflict with Congress’ expressed intent.” Rust v. Sullivan, 500 U.S. 173, 184, 111 S.Ct. 1759, 114 L.Ed.2d 233 (1991).
Here, Congress directed NMFS to promulgate a fishery management plan and implementing regulations that have the force of law through notice-and-comment rulemaking. See 16 U.S.C. § 1854. Accordingly, we must defer to NMFS’s construction of ambiguous statutory provisions authorizing such plans so long as its interpretation is reasonable.
A
We first consider Glacier’s argument that it is not subject to cost recovery because NMFS is authorized to collect a fee only from “limited access privilege holders,” 16 U.S.C. § 1853a(e)(2). Glacier claims that it is not such a holder for two reasons: first, because a catcher-processor coop permit is not a limited access privilege, and second, because even if a coop permit is a limited access privilege, PWCC holds the permit, not Glacier.
Under the Chevron framework, we begin with the question whether Congress has addressed the precise question before us, that is, whether a “limited access privilege” may include a permit such as the catcher-process coop permit, as set forth in Amendment 20 and the implementing regulations.
The Magnuson-Stevens Act defines a “limited access privilege” as “a Federal permit, issued as part of a limited access system under section 1853a of this title to harvest a quantity of fish .., representing a portion of the total allowable catch of the fishery that may be received or held for exclusive use by a person,” 16 U.S.C. § 1802(26)(A), and “includes an individual fishing quota,” id. § 1802(26)(B). Thus, the definition of a “limited access privilege” makes clear that the IFQ permits used for participants in the shoreside sector are limited access privileges. But the definition does not resolve whether a coop permit, which is not issued directly to harvesters, but to a coop to receive and manage a catch allocation on behalf of its members, is a “permit ... .to harvest a quantity of fish.” Nor does any other part of the statute answer this question. Although the Act has delineated the contours of “limited access privilege programs” in a detailed statutory framework, id. § 1853a, including the requirements for a limited access privilege, id. § 1853a(c), and the characteristics of such a privilege, id. § 1853a(f), we have not found, and the parties have not identified, any provisions in the Magnuson-Stevens Act that specifically address whether a limited access privilege could include a permit issued to a coop along the lines set forth in NMFS’s regulations. We therefore conclude that Congress has not directly spoken to this issue.
Accordingly, we move to the second step of the Chevron analysis and ask whether NMFS’s construction of “limited [1122]*1122access privilege” as including the coop permit defined in NMFS’s regulations was reasonable. According to the regulations, the “coop program” is “a limited access program that applies to vessels in the C/P sector of the Pacific whiting at-sea trawl fishery and is a single voluntary coop.” 50 C.F.R. § 660.160(a). The “coop permit” is defined as a permit that “conveys a conditional privilege to an eligible coop entity to receive and manage a coop’s allocation of designated species and species groups.” Id. § 660.25(e)(2). A coop is a group of participants that acts both as a group and individually: the regulations define a coop as “a harvester group” that includes participants in the catcher-processor sector with limited entry permits who “voluntarily form a coop and who manage the catcher/processor-specified allocations through private agreements and contracts,” id. § 660.111, and the harvester group and its members are jointly and severally responsible for compliance with the coop permit and coop program, id. § 660.160(b)(2)(iii). In order to harvest fish, participants in the catcher-processor sector must be members of the coop, and the coop must have a coop permit. Id. § 660.160(a), (b)(2)(i)(A). Once the coop permit has been obtained, coop members may engage in catching the catcher-processor sector’s allocation during the season. Id. § 660.160(d)(l)(iv). In sum, as NMFS explained in proposing the regulations, the “coop permit” registers “the coop and its associated members to harvest and process in the sector.” 75 Fed. Reg. 53,380-01, 53,392.
We conclude that NMFS could reasonably interpret a “federal permit ... to harvest a quantity of fish” as including the coop permit. Taking the regulations as a whole, a coop permit is a federal permit issued to a harvester group, 50 C.F.R. § 660.111, which is necessary (but not sufficient) for the group members to harvest the fish allocated to the catcher-processor sector, id. § 660.160(b)(2)(i)(A), and which allows the group members to apportion the harvest of fish amongst themselves through private agreement, id. § 660.111. Because the coop permit allows the group to catch the portion of the fish allocated to the catcher-processor sector, it can reasonably be called a “federal permit ... to harvest a quantity of fish,” 16 U.S.C. § 1802(26). That participants in the fishery obtain their right to catch fish by means of a coop permit rather than an individual permit does not change the underlying nature of the permit. Because NMFS’s determination was a plausible construction of the statute, we defer to the agency’s interpretation.
Glacier argues that because the coop permit is defined as “a conditional privilege to an eligible coop entity to receive and manage a coop’s allocation of designated species and species groups,” 50 C.F.R. § 660.25(e)(2), rather than expressly stating that it is a permit to “harvest” fish, the coop permit is not a “limited access privilege.” And, Glacier adds, a coop itself cannot “harvest” fish in the ordinary sense of the word because it does not have the physical or legal means to catch fish, so the coop permit cannot be a permit to “harvest” fish. We do not dispute that “harvest” should be given its ordinary meaning and in this context generally means to catch fish. But we reject Glacier’s argument because it fails to consider the nature of a coop and a coop permit in the context of the regulation as a whole, which establishes a coop permit as an authorization allowing the members of a harvester group to catch a certain quantity of fish.7 Congress gave NMFS authority to [1123]*1123develop a limited access privilege program, and “agencies are better equipped to make” the policy choices for doing so than are the courts. Brand X, 545 U.S. at 980, 125 S.Ct. 2688. We therefore defer to NMFS’s determination that a coop permit is a permit allowing a harvesting group to harvest fish.
Glacier also argues that that NMFS cannot deem a coop permit to be a “limited access privilege” because the catcher-processor coop permit lacks one of the characteristics of a limited access privilege set forth in § 1853a(f). The missing characteristic, according to Glacier, is that a limited access privilege is “issued for a period of not more than 10 years that — (1) will be renewed before the end of that period, unless it has been revoked, limited, or modified as provided in this subsection.” 16 U.S.C. § 1853a(f). Under NMFS’s regulations, an application for a coop permit must be submitted to NMFS between February 1 and March 31 of each year and it expires on December 31 of the year in which it was issued. 50 C.F.R. § 660.160(d)(1)(h).8 Glacier claims that because the coop permit lacks the characteristic of renewability, it is not a limited access privilege.
This argument misses the point. Rather than raise a question as to whether it was unreasonable for NMFS to construe a “limited access privilege” as including the type of “coop permit” defined in NMFS’s regulations, this argument raises a different question: whether the regulation setting forth the coop permit’s renewal requirements is a reasonable interpretation of the statutory renewal requirements in § 1853(a)(f). The answer to this question sheds no light on whether a coop permit is a “limited access privilege.” Here, for instance, if NMFS fails to renew the coop permit before the end of the one-year period, Glacier could raise the argument that NMFS’s actions are inconsistent with § 1853a(f), which requires renewal of a permit “unless it has been revoked, limited or modified” pursuant to § 1853a(f)(l). If Glacier prevailed on this claim, NMFS could have the obligation to revise or interpret the limited access privilege to meet statutory requirements. But regardless of Glacier’s success on this dlaim, the coop permit would still be “a Federal permit, issued as part of a limited access system under section 1853a of this, title to harvest a quantity of fish ... representing a portion of the total allowable catch of the fishery that may be received or held for exclusive use by a person,” 16 U.S.C. § 1802(26), and therefore would constitute a limited access privilege.' Thus, Glacier’s renewability argument does not change our conclusion that NMFS could reasonably determine the coop permit to be a limited access privilege.
We turn next to Glacier’s argument that, even if the catcher-processor coop permit is a limited access privilege, NMFS is authorized to collect a fee only from “limited access privilege holders,” 16 U.S.C. § 1853a(e)(2) (emphasis added), and Glacier is not a “holder” of thgt permit because it was issued only to PWCC.
We again begin by determining whether Congress has addressed the precise question before us, which is whether each indi[1124]*1124vidual member of a coop, defined as a “harvester group,” 50 C.F.R- § 660.111, can be deemed a “holder” of the Federal permit issued to that group. The word “holder” is not defined in the statute. Nor is there “an unambiguous common sense meaning of the word” that answers the question here. See Arizona Health Care Cost Containment Sys. v. McClellan, 508 F.3d 1243, 1249 (9th Cir. 2007). Although the term “holder” as it is used here means “possessor,” Webster’s Third New International Dictionary 1079 (2002), this definition does not resolve our issue; rather, it merely raises the same question: whether a member of a group may be deemed to be a “possessor” of a group permit. We therefore conclude that Congress has not directly answered the precise question at issue.
Turning to the second step of Chevron, we ask whether NMFS’s determination that each member of the harvester group that received the coop permit can be treated as a “limited access privilege holder” is a reasonable construction of § 1853a(e)(2). We conclude that it is. As we have explained, the regulations define the coop as a “harvester group,” which is treated for regulatory purposes as both the group and the individual members.' Because each member of the group exercises the group’s rights under the coop permit to harvest fish, see 50 C.F.R. § 660.160(b), (d)(l)(iv), and each member is jointly and severally responsible for compliance with the permit, see id. § 660.160(b)(2)(iii), it is both fair and sensible to regard each member of the group as a joint holder of the privilege. Accordingly, we defer to NMFS’s reasonable construction of “holder” in § 1853a(e)(2) as being applicable to each member of a coop that has been issued a limited access privilege. Id. § 660.115(d)(2).
Because the catcher-processor coop permit is a limited access privilege, and Glacier can reasonably be said to be a “holder” of that permit, NMFS had the authority to collect a fee from the individual members of the coop, including Glacier. The district court therefore did not err in granting summary judgment to NMFS on this issue.
B
We turn next to Glacier’s argument that NMFS did not apply the Council’s cost accounting methodology, which Glacier argues violated 16 U.S.C. §§ 1853a(e), 1854(d)(2).
The Magnuson-Stevens Act requires a council that elects to establish a limited access privilege program to “(1) develop a methodology and the means to identify and assess the management, data collection and analysis, and enforcement programs that are directly related to and in support of the program,” and “(2) provide ... for a program of fees paid by limited access privilege holders that will cover the costs of management, data collection and analysis, and enforcement activities.” Id. § 1853a(e). After the council develops the methodology and provides the program, NMFS may accept or reject the program proposed by the council, id. § 1854(b), but it does not have the authority to develop a methodology itself.
The Pacific Council developed a methodology for collecting costs “directly related to and in support of the program,” id. § 1853a(e)(l), and prepared a report with its recommendation. In its report, the Council recommended recovering only “incremental costs, i.e. those costs that would not have been incurred but for” the trawl rationalization program. It defined “[a]ctual incremental costs” to mean “those net costs that would not have been incurred but for the implementation of the trawl rationalization program including additional costs for new requirements of the program and reduced trawl sector related [1125]*1125costs resulting from efficiencies as a result of the program.” “Net costs” were to be interpreted pursuant to Appendix B, which provided some guidance and examples for determining how the cost of an employee’s time changes with and without a trawl rationalization program.
NMFS followed the Council’s recommendation in its regulations. The final regulations required each participant in the fishery to pay as an annual fee a percentage “of the ex-vessel value of fish harvested by sector under the trawl rationalization program,” but no more than 3 percent under 16 U.S.C. § 1854(d)(2)(B). 50 C.F.R. § 660.115(a). The regulations required NMFS to calculate the fee percentage each year using the formula, (DPC/V) x 100. Id. § 660.115(b)(1). It defined DPC, or direct program costs, as “the actual incremental costs for the previous fiscal year directly related to the management, data collection, and enforcement of each sector (Shorebased IFQ Program, MS Coop Program, and C/P Coop Program).” Id. § 660.115(b)(l)(I). The regulations adopted the council’s definition of “actual incremental costs”: the “net costs that would not have been incurred but for the implementation of the trawl rationalization program, including additional costs for new requirements of the program and reduced trawl sector related costs resulting from efficiencies as a result of the program.” Id. NMFS therefore sought to collect only the “incremental costs” as the Council defined them, the costs that would not have been incurred but for the implementation of the rationalization program.9 Because NMFS’s regulations implemented the Council’s recommendations, the regulations are consistent with the statutory requirement.
Glacier argues that the cost recovery regulations violated § 1858a(e) because NMFS did not use the “with and without” approach in Appendix B. But Appendix B’s “with and without” approach is merely an example of how to calculate “net costs that would not have been incurred but for the implementation of the trawl rationalization program.” In other words, Appendix B demonstrated that the net costs in this context are equal to costs “with” the implementation of the trawl rationalization program less costs “without” that program. Because NMFS’s regulations include the Council’s definition of incremental costs as net costs with and without the trawl rationalization program, NMFS complied with § 1853a(e), and we reject Glacier’s argument to the contrary.
C
Finally, we address Glacier’s argument that NMFS’s calculation of the 2014 cost recovery fee was inconsistent with NMFS’s own cost recovery regulations because NMFS did not determine the “actual” costs “directly related to” the program and failed to account for any efficiencies gained as a result of the program. We generally defer to the agency’s interpretation of its own regulations unless “plainly erroneous or inconsistent with the regulation.” Bassiri v. Xerox Corp., 463 F.3d 927, 930 (9th Cir. 2006) (citing Auer v. Robbins, 519 U.S. 452, 461, 117 S.Ct. 905, 137 L.Ed.2d 79 (1997)); see also Chase Bank USA, N.A. v. McCoy, 562 U.S. 195, 207-08, 131 S.Ct. 871, 178 L.Ed.2d 716 (2011).
The regulations require NMFS to begin by determining the “additional costs” of [1126]*1126the trawl rationalization program attributable to each sector. 50 C.F.R. § 660.115(b)(l)(I).10 Accordingly, NMFS was obliged to develop a reasonable method for determining the actual additional costs of the trawl rationalization program (over and above existing costs related to groundfish fishery management) as well as a reasonable method for determining which of these additional costs was directly attributable to each of the three sectors. Moreover, the regulations required NMFS to develop a method to evaluate whether it has gained any “efficiencies” and reduced costs as a result of implementing the coop permit program. Id.
A review of the record discloses that NMFS did not develop or apply such reasonable methods. First, NMFS did not have a systematic method for determining the number of employee hours that were spent on the trawl rationalization program over and above the amount that had previously been spent on the groundfish fishery. Instead, NMFS made rough approximations. It began with the presumption that all employee time should be treated as the incremental cost of the trawl rationalization program because it had hired many new employees solely for work on the trawl rationalization program. NMFS acknowledged that this presumption was not wholly justified, because NMFS had retained some employees who had been working on the same tasks before and after, the implementation of the program. To address this problem, NMFS contended that time spent by existing employees was offset by NMFS’s decision not to obtain cost recovery for certain overhead costs. This approach of using roughly estimated offsets is ’ not consistent with NMFS’s regulatory obligation to determine “the actual incremental costs” of developing the coop program. 50 C.F.R. § 660.115(b)(l)(I) (emphasis added).
Nor did NMFS develop a reasonable system for separating time spent on the trawl rationalization program from time spent on other projects. NMFS asked employees to track the hours they spent working on trawl rationalization matters related to each of the groundfish fishery sectors (using a code assigned to each of these sectors), plus any hours they spent working on general matters not specifically attributable to any sector (using a code for “general work”). The record shows that managers used rough estimates and deducted hours on an ad hoc basis. For instance, over two thousand hours were deleted because they “seem[ed] high.” A manager deleted one employee’s time entirely because it turned out she had, not spent any time on the trawl rationalization program. NMFS also had problems with the “general” category. A manager deleted one employee’s time in the general category because that employee had not spent any time on general work, and reduced another’s time by 75 percent because the manager learned, based on personal communications with the employee, that the employee had coded the time he spent on projects other than the trawl rationalization program as “general” work. Again, such a haphazard approach to estimating the number of employee hours actually spent on the trawl rationalization program is inconsistent with NMFS’s regulatory obligation to calculate the actual incremental costs directly related to the program.
[1127]*1127Moreover, NMFS did not allocate time in the general category according to the sector that required the most employee time; rather, NMFS divided the cost of the general time (including employee leave time) evenly among the three sectors even though employee costs of the catcher-processor sector amounted to only 1.4% of the sector-specific employee costs.11 The record shows that the vast majority of the costs assigned to the catcher-processor sector is from the “general” category: of the $176,460 in costs allocated to the catcher-processor sector, only $25,807 was for time spent on the catcher-processor sector, while the other $150,658 constituted one-third of the “general” time spent on the trawl rationalization program as a whole. Distributing general costs like employee leave time evenly across the three sectors is again inconsistent with calculating the “actual incremental costs ... directly related to ... each sector.” 50 C.F.R. § 660.115(b)(l)(I).
Finally, NMFS apparently made no attempt to calculate the “reduced trawl sector related costs resulting from efficiencies” in the program as it was required to do under 50 C.F.R. § 660.115(b)(l)(I).
Based on our review of the record, NMFS did not properly determine the “actual incremental costs” that were “directly related to the management, data collection, and enforcement of each sector” for assessment on the members of the catcher-processor sector. 50 C.F.R. § 660.115(b)(l)(I). We therefore conclude that NMFS’s calculation of the 2014 cost recovery fee of the catcher-processor sector was inconsistent with NMFS’s own regulations. We reverse the district court’s grant of summary judgment to the extent it upheld NMFS’s fee calculation and remand to the agency to re-determine that fee in accordance with its regulations.
Each party shall bear its own costs on appeal.
AFFIRMED in part, REVERSED AND REMANDED in part.