United States Court of Appeals
         FOR THE DISTRICT OF COLUMBIA CIRCUIT



Argued April 16, 2019              Decided August 20, 2019

                        No. 18-5154

            AMERICAN BANKERS ASSOCIATION,
                      APPELLEE

                            v.

        NATIONAL CREDIT UNION ADMINISTRATION,
                     APPELLANT


                Consolidated with 18-5181


       Appeals from the United States District Court
               for the District of Columbia
                   (No. 1:16-cv-02394)


     Daniel Aguilar, Attorney, U.S. Department of Justice,
argued the cause for Appellant-Cross-Appellee. With him on
the briefs was Mark B. Stern, Attorney.

    Allison Jones Rushing was on the brief for amici curiae
Credit Union National Association, et al. in support of
Appellant-Cross-Appellee. Nicholas G. Gamse entered an
appearance.
                               2
   Steven D. Gordon was on the brief for amici curiae State
Bankers Associations in support Appellee-Cross-Appellant
American Bankers Association.

    Robert A. Long Jr. argued the cause for Appellee-Cross-
Appellant. With him on the briefs were Andrew J. Soukup,
Philip Levitz, and Lauren Moxley.

    Before: HENDERSON, PILLARD, and WILKINS, Circuit
Judges.

    Opinion for the Court filed by Circuit Judge WILKINS.

     WILKINS, Circuit Judge: Longstanding principles of
administrative law teach us to give federal agencies breathing
room when they make policy and “resolv[e] the struggle
between competing views of the public interest.” Chevron,
U.S.A., Inc. v. Nat. Res. Def. Council, Inc., 

, 866
(1984). And because many policy decisions merge with legal
ones, Chevron requires us frequently to sustain agency
interpretations of certain federal statutes. Congress often
expects agencies, with their political accountability, “bod[ies]
of experience[,] and informed judgment,” to make sound
interpretive choices “with the force of law.” United States v.
Mead Corp., 

, 227, 229 (2001) (citation omitted).

     Congress expressly tasked the National Credit Union
Administration (NCUA) with making such choices in defining
the reach of federal credit unions. Since the Great Depression,
Congress has maintained a “system of federal credit unions
that . . . provide credit at reasonable rates” and banking
services to “people of ‘small means.’” First Nat’l Bank & Tr.
Co. v. NCUA (First Nat’l Bank I), 

, 1274 (D.C.
Cir. 1993) (citation omitted), aff’d, 

(1998).
Although a private bank may solicit and welcome customers
                                3
from anywhere, Congress has limited whom these federal
financial institutions may serve.        For instance, certain
institutions called “community credit unions” may cover
individuals and entities only within a preapproved
geographical area. The credit union will not receive a federal
charter (and thus cannot start operations) unless it first proffers
a geographical coverage area and the NCUA accepts the
proposal. Congress explicitly assigns the agency the task of
creating vetting standards.

     Exercising its expressly delegated power, the NCUA has
promulgated a final rule that makes it easier for community
credit unions to expand their geographical coverage and thus to
reach more potential members. Representing competitors to
the credit unions, the American Bankers Association
(Association) has challenged the NCUA’s new rule as neither
“in accordance with law” nor within “statutory jurisdiction.” 5
U.S.C. § 706(2)(A), (C). The District Court vacated significant
portions of the rule, deeming them to be based on unreasonable
agency interpretations of the Federal Credit Union Act (Act),
Pub. L. No. 73-467, 48 Stat. 1216 (1934) (codified as amended
at 12 U.S.C. §§ 1751 to 1795k). See Am. Bankers Ass’n v.
NCUA, 

, 61, 69-70 (D.D.C. 2018).

     We appreciate the District Court’s conclusions, made after
a thoughtful analysis of the Act. But we ultimately disagree
with many of them. In this facial challenge, we review the rule
not as armchair bankers or geographers, but rather as lay judges
cognizant that Congress expressly delegated certain policy
choices to the NCUA. After considering the Act’s text,
purpose, and legislative history, we hold the agency’s policy
choices “entirely appropriate” for the most part. 

. We therefore sustain the bulk of the rule. Still, we
do not rubber-stamp this regulation. We remand, without
vacating, one portion for further consideration of the
                               4
discriminatory impact it might have on poor and minority
urban residents.

                               I.

                              A.

     The nation’s credit unions started in the early twentieth
century “as a populist mechanism designed to empower
farmers against bad loans.” Mehrsa Baradaran, How the Poor
Got Cut Out of Banking, 62 EMORY L.J. 483, 500 (2013).
Walloped by crop failures and the Great Depression, farmers
seeking credit became not only increasingly suspicious of
traditional bankers, who “disregard[ed]” poor individuals and
stayed in the big cities, but also fearful of loan sharks, “who
would extract ‘up to a thousand percent’ in interest rates.” 

(quoting 80 CONG. REC. 6752 (1936) (statement of
Rep. Lundeen)). The farmers thus began to build their own
credit networks.

     In a national grassroots campaign, farmers created
localized, non-profit “credit groups” collecting funds from and
loaning small sums to one another at low interest rates. See 

The success of any such self-help institution
“hinge[s] on the interpersonal dynamics of its members:
Lenders must be able to evaluate the ability and willingness of
potential borrowers to pay back their loans and borrowers must
feel obligated to pay back those loans.” Wendy Cassity, Note,
The Case for a Credit Union Community Reinvestment Act, 100
COLUM. L. REV. 331, 337 (2000); see also First Nat’l Bank &
Tr. Co. v. NCUA (First Nat’l Bank II), 

, 526 (D.C.
Cir. 1996), aff’d, 

(1998).

    By 1934, individuals had organized about 3,000 local
credit unions, with about 750,000 members. See 80 CONG.
                               5
REC. at 6753. Recognizing the success of credit unions at the
state level, Congress created a federal system that year by
passing the Act. Legislators worried that “usurious money
lending . . . obviously destroy[ed] vast totals of buying power
[once held by] . . . the average worker.” H.R. REP. NO. 73-
2021, at 1-2 (1934); see also S. REP. NO. 73-555, at 1 (1934).
Congress touted the Act’s ability to “make more available to
people of small means credit for provident purposes.” H.R.
REP. NO. 73-2021, at 1; see also S. REP. NO. 73-555, at 1.

     Credit unions multiplied over the ensuing decades. By
1970, Congress created an independent agency to supervise
federal credit unions: the NCUA. See Pub. L. No. 91-468, 84
Stat. 994 (1970) (codified as amended in scattered sections of
12 U.S.C.); see also Swan v. Clinton, 

, 974 (D.C.
Cir. 1996) (noting that Congress “entrusted” the agency with
“the responsibility of overseeing” federal credit unions).
Legislators thought that the agency would be “more responsive
to the needs of credit unions” and would “provide more flexible
and innovative regulation” than prior government agencies,
which did not have federal credit unions as their sole focus. S.
REP. NO. 91-518, at 3 (1969).

     The NCUA faced its first major crisis at the end of the
1970s. After years of economic decline in several industrial
sectors, federal credit unions tied to those business sectors
began to suffer. The resulting liquidation of numerous credit
unions “threaten[ed] ‘the safety and soundness of the federal
credit union system.’” 

(footnote
omitted). Reacting to the emergency, the NCUA in 1981
promulgated a groundbreaking rule that loosened a major size
limitation on certain federal credit unions.          Almost
immediately, those financial institutions grew in membership.
                               6
     Meanwhile, credit unions became “caught up in the
broader changes in banking and faced internal as well as
external pressure to compete with [private] banks and seek
higher profits.” 

. Unlike credit
unions, private, for-profit banks were “owned by equity holders
who may not necessarily be customers (depositors or
borrowers),” and they did “not have similar membership and
commercial lending restrictions” as credit unions. DARRYL E.
GETTER, CONG. RESEARCH SERV., IF11048, INTRODUCTION TO
BANK REGULATION: CREDIT UNIONS AND COMMUNITY BANKS:
A COMPARISON 1 (2018). To remain viable, credit unions
“started to focus on attracting more customers and expanding
the industry.” 

. As part of that
strategy, many consolidated through mergers. And private
banks soon treated credit unions as serious competitors,
seeking to curb their growth. See NCUA v. First Nat’l Bank &
Tr. Co. (First Nat’l Bank III), 

, 485 (1998); First
Nat’l Bank 

.

     In 1998, the banking industry successfully challenged as
contrary to the Act the 1981 rule that had eased size limitations
for certain federal credit unions. See First Nat’l Bank 

. Congress swiftly responded. In less than six
months, legislators amended the Act, superseding the holding
in First National Bank III, loosening size limitations on certain
federal credit unions, and adding other reforms. See Credit
Union Membership Access Act, Pub. L. No. 105-219, 112 Stat.
913 (1998) (codified as amended in scattered sections of 12
U.S.C.). Partly because of the 1998 amendments and related
NCUA regulations, credit unions continued to merge and grow
in membership. Now, more than 61 million customers perform
their banking services at about 3,400 federal credit unions. See
2018 NAT’L CREDIT UNION ADMIN. ANN. REP. 192.
                               7
                               B.

     Federal credit unions pool funds from – and give loans
to – their members and other credit-union entities. 12 U.S.C.
§ 1757(5), (6). A credit union’s members, whether individual
or corporate, must come from the credit union’s membership
“field,” 

which is based on a shared occupation,
association, or geographical area. Members receive regular
dividends. 

Congress has shielded federal credit
unions from federal corporate income taxes and most state and
local taxes, but members must pay taxes on their dividends.
See JAMES M. BICKLEY, CONG. RESEARCH SERV., 97-548 E,
SHOULD CREDIT UNIONS BE TAXED? 3-5 (2005).

     To create a federal credit union, at least seven individuals
must present a proposed charter and pay a fee to the NCUA.
See MICHAEL P. MALLOY, BANKING LAW & REGULATION
§ 2.04 (2d ed. 2019). In the application, the organizers must
pledge to deposit funds for shares in the institution and must
describe the credit union’s proposed membership field. 12
U.S.C. § 1753(3), (5). The NCUA must approve the charter
before the institution may start. See 

The agency
will complete an “appropriate investigation” and determine the
“general character and fitness” of the organizers, the
“economic advisability of establishing” the credit union, and
the “conform[ity]” of proposal details with the Act. 

governs two types of federal credit unions:
“common-bond” credit unions and “community” credit unions.
See 

This case deals with the latter category. The
1934 version of the Act required a community credit union’s
membership field to reflect a particular geographical area – to
wit, “a well-defined neighborhood, community, or rural
district.” § 9, 48 Stat. at 1219. As amended in 1998, the Act
provides that membership for a community credit union “shall
                               8
be limited to . . . [p]ersons or organizations within a well-
defined local community, neighborhood, or rural district.” 12
U.S.C. § 1759(b) (emphasis added). The 1998 version calls on
the NCUA to “prescribe, by regulation, a definition for the term
‘well-defined local community, neighborhood, or rural
district.’” 

Thus, under the new regime,
individuals seeking to organize a new community credit union
(or alter an existing one) must commit to serving members
within the NCUA’s contemporaneous definition of “local
community, neighborhood, or rural district.” See S. REP. NO.
105-193, at 4, 8 (1998); H.R. REP. NO. 105-472, at 21 (1998).
As part of their application to the NCUA, they must provide a
proposed description of the precise geographical area that the
credit union would serve.

    Since 1998, there has been “dramatic growth” in the
number of community credit unions.              U.S. GOV’T
ACCOUNTABILITY OFF., GAO-07-29, CREDIT UNIONS:
GREATER TRANSPARENCY NEEDED ON WHO CREDIT UNIONS
SERVE AND ON SENIOR EXECUTIVE COMPENSATION
ARRANGEMENTS 4 (2006). Despite a 11-percent drop in the
number of federal credit unions from 2000 to 2005, community
credit unions doubled to 1,115. 

12. Meanwhile, the
amount of assets in community credit unions quadrupled to
$104 billion. 

                              C.

    On December 7, 2016, the NCUA amended its
membership-field rules for community credit unions. See
Chartering and Field of Membership Manual, 81 Fed. Reg.
88,412 (Dec. 7, 2016). Several changes rely on two terms
devised by the Office of Management and Budget (OMB) and
based on data collected by the Census Bureau (Census): “Core
Based Statistical Areas” and “Combined Statistical Areas.”
                              9

     The OMB has designated numerous regions around the
country as Core Based Statistical Areas, which comprise at
least one urban cluster, or core, of 10,000 or more people and
adjacent counties with substantial commuting ties to that core.
See U.S. CENSUS BUREAU, GEOGRAPHICAL PROGRAM,
GLOSSARY, https://www.census.gov/programs-surveys/geogra
phy/about/glossary.html. In layman’s terms, a Core Based
Statistical Area is a city or town and its suburbs.

     Meanwhile, a Combined Statistical Area is a conglomerate
of two or more adjoining Core Based Statistical Areas, each of
which has substantial commuting ties with at least one other
Core Based Statistical Area in the group. 

Combined Statistical Area is a regional hub with urban centers
connected by commuting patterns. Combined Statistical Areas
may “reflect broader social and economic interactions, such as
wholesaling, commodity distribution, and weekend recreation
activities.” OFFICE OF MGMT. & BUDGET, EXEC. OFFICE OF THE
PRESIDENT, OMB BULL. NO. 15-01, REVISED DELINEATIONS OF
METROPOLITAN STATISTICAL AREAS, MICROPOLITAN
STATISTICAL AREAS, AND COMBINED STATISTICAL AREAS, AND
GUIDANCE ON USES OF THE DELINEATIONS OF THESE AREAS
app. 2-3 (2015).

     Relevant here, the 2016 rule made two changes to the
NCUA’s definition of the term “local community” under
§ 1759(b)(3) and one to that of “rural district.” The changes
affect what proposed membership areas satisfy the
geographical limitation imposed by the Act.

    The first change to the “local community” definition
involves Combined Statistical Areas. A proposed area
qualifies as a local community if it encompasses the whole or
a portion of a Combined Statistical Area and does not exceed a
                                10
designated population limit. See 81 Fed. Reg. at 88,440. The
NCUA has set that cap at 2.5 million people.

     The second change involves Core Based Statistical Areas.
The parties agree that all or part of a Core Based Statistical
Area may qualify as a local community so long as it does not
exceed the population limit. But since 2010, the NCUA
required such a membership area to include the urban core.
The new rule no longer requires that the core be included in the
local community that a credit union proposes to serve. See 

88,440.

     As for the “rural district” definition, the new rule increases
the population cap for valid rural districts from 250,000 people
(or 3 percent of the population of the state where most eligible
residents are located) to 1 million people. See 

88,440. The new population limit works with two other
constraints set by the rule: (1) an outer geographical limit on
how far a rural district may extend past the borders of the credit
union’s headquarters state; and (2) a requirement either that
most eligible residents reside in Census-designated rural areas,
or that the population density of the proposed district equals
100 or fewer people per square mile. See 

                                D.

     On the day the NCUA published the rule, the Association
filed this injunctive and declaratory action in the District Court.
The Association claimed that the three changes described
above were not only arbitrary and capricious under the
Administrative Procedure Act (APA), Pub. L. No. 79-404, 60
Stat. 237 (1946) (codified as 5 U.S.C. §§ 701-06), but also
unreasonable and entitled to no deference under Chevron. The
agency and Association filed cross-motions for summary
                                11
judgment. On March 29, 2018, the District Court granted both
motions in part and denied them in part.

     The court made three relevant holdings. First, it rejected
as unreasonable the qualification of certain Combined
Statistical Areas as local communities. See Am. Bankers 

. Second, it sustained as well-reasoned
the elimination of the core requirement from the Core Based
Statistical Area a credit union proposes to serve as its local
community. 

Third, it rejected as unreasonable the
increased population cap for rural districts. 

(It
also sustained a separate portion of the rule, which the
Association does not challenge here.)

    The NCUA and Association timely appealed. We have
appellate jurisdiction under 28 U.S.C. § 1291.

                                II.

    At the outset, we must assure ourselves of our subject
matter jurisdiction over the appellate proceeding. See, e.g.,
United States v. Gooch, 

, 1277 (D.C. Cir. 2016).

     In their original briefing, the parties failed to apprise us of
a rule that was promulgated while this appeal was pending and
that changed membership-field requirements for community
credit unions. See Chartering and Field of Membership, 83
Fed. Reg. 30,289 (June 28, 2018). The 2018 rule eliminated
the portion of the 2016 rule allowing Combined Statistical
Areas to qualify as local communities. Compare 12 C.F.R. pt.
701, app. B, ch. 2 § V.A.2 (2018), with 

2018
rule preamble did not specifically discuss the removal but
concluded that “[a]ny modification in th[e] final rule is
consistent with the District Court decision” in this action. 83
Fed. Reg. at 30,291.
                               12

     “Under the mootness doctrine, we cannot decide a case if
‘events have so transpired that the decision will neither
presently affect the parties’ rights nor have a more-than-
speculative chance of affecting them in the future.’” Reid v.
Hurwitz, 

, 832 (D.C. Cir. 2019) (citation omitted).
The same principle applies to individual claims. See Tucson
Med. Ctr. v. Sullivan, 

, 977 (D.C. Cir. 1991).
Accordingly, if a rule under review (or a portion of it) is
superseded or amended during an appeal, the proceeding (or
relevant part) might be moot. See, e.g., Am. Bankers Ass’n v.
NCUA, 

, 274 (D.C. Cir. 2001). We therefore
requested supplemental briefing on the issue of appellate
jurisdiction.

     Based on the government’s submission and
representations at oral argument, we hold that the portion of the
appeal related to Combined Statistical Areas is not moot.
“[T]he mere power to [reinstitute] a challenged law is not a
sufficient basis on which a court can conclude that” a challenge
remains live. Nat’l Black Police Ass’n v. District of Columbia,

, 349 (D.C. Cir. 1997). The government has stated
that, if we reverse the relevant part of the District Court’s
decision, all three members of the NCUA’s board intend to
reinstitute the Combined Statistical Area portion of the 2016
rule. See Decl. of Michael McKenna ¶ 3, ECF No. 1781123;
Oral Arg. Recording 11:33-48. Accordingly, City of Mesquite
v. Aladdin’s Castle, Inc., 

(1982), governs this
case. In Aladdin’s Castle, the Supreme Court deemed the
controversy live in part because the government had announced
“an intention” to restore the rule under challenge if the lower-
court decision were vacated. 

& n.11; see also Am.
Bankers 

. The NCUA’s submission and
representations evince such an intention here, and the
Association – which bears the “heavy burden” of proving
                               13
mootness, Friends of the Earth, Inc. v. Laidlaw Envtl. Servs.
(TOC), Inc., 

, 189 (2000) (citation
omitted) – offers no evidence to the contrary.

     The Association attempts to distinguish Aladdin’s Castle
on two grounds, but neither sways us. First, the Association
notes that the city government in Aladdin’s Castle said it would
reenact “precisely the same” law, 

, but that
in this case any future notice-and-comment proceedings might
produce a different “local community” definition, perhaps not
even relying on Combined Statistical Areas. After all, the
agency must keep a “flexible and open-minded attitude” during
the process. See Nat’l Tour Brokers Ass’n v. United States, 

, 902 (D.C. Cir. 1978). We grant that commenters
might convince the NCUA to change its mind; mann tracht,
und Gott lacht. But that strikes us as speculative as public input
convincing Mesquite legislators to enact a different law. We
do not see Aladdin’s Castle turning on such conjecture.
Instead, we see a live dispute because there is “no certainty”
that the NCUA will forego reinstating the same Combined
Statistical Area definition. Aladdin’s 

;
see also Trinity Lutheran Church of Columbia, Inc. v. Comer,

, 2019 n.1 (2017) (holding that the court must
find “absolutely clear that the allegedly wrongful behavior
could not reasonably be expected to recur” (emphasis added)
(quoting Friends of the 

)). After all, the
NCUA continues to defend the definition here. See Knox v.
Serv. Emps. Int’l Union, Local 100, 

, 307 (2012).

     Second, the Association observes that both sides in
Aladdin’s Castle urged the Supreme Court to treat their dispute
as live. In contrast, only the NCUA seeks to proceed here; the
Association would prefer to wait until the agency reinstitutes
the rule. But the existence or absence of jurisdiction does not
turn on which parties challenge or defend it. Cf. Bender v.
                              14
Williamsport Area Sch. Dist., 

, 541 (1986).
Because the NCUA remains bound by the lower-court
judgment, the present injury renders irrelevant the
Association’s preference. “Jurisdiction existing,” our duty to
decide the appeal “is ‘virtually unflagging.’”          Sprint
Commc’ns, Inc. v. Jacobs, 

, 77 (2013) (citation
omitted).

     In short, we may review the challenge to the rule change
involving Combined Statistical Areas. We see no jurisdictional
issues with the rest of the appeal. We thus turn to the merits.

                              III.

    We review de novo the District Court’s rulings on
summary judgment. See Loan Syndications & Trading Ass’n
v. SEC, 

, 222 (D.C. Cir. 2018). We review the
administrative record and give “no particular deference” to the
District Court’s views. Oceana, Inc. v. Ross, 

, 860
(D.C. Cir. 2019) (citation omitted).

     The APA governs this suit. In relevant part, the statute
provides that we “decide all relevant questions of law” and
“interpret . . . statutory provisions.” 5 U.S.C. § 706. We
ordinarily set aside agency actions that are either “arbitrary,
capricious, an abuse of discretion, or otherwise not in
accordance with law,” 

or “in excess of
statutory jurisdiction, authority, or limitations, or short of
statutory right,” 

    We review the agency rule in accordance with the familiar
Chevron doctrine, a two-prong test for determining whether an
agency “has stayed within the bounds of its statutory authority”
when issuing its action. City of Arlington v. FCC, 

, 297 (2013) (emphasis omitted). At the first step, we
                             15
determine “whether Congress has directly spoken to the precise
question at issue,” and we “give effect” to any “unambiguously
expressed intent.” 

& n.9.

     If we glean no such unambiguous intent, we turn to the
second step and determine “whether the agency’s answer” to
the question “is based on a permissible construction of the
statute.” 

By arriving at the second step, we have
concluded that Congress either explicitly or implicitly
delegated to the agency the lawmaking authority to clarify the
statute. We presume that Congress would not authorize the
promulgation of an “[im]permissible construction.” 

. Accordingly, we will set aside agency actions
based on such a construction, because they are either “not in
accordance with law” or “in excess of statutory jurisdiction.”
5 U.S.C. § 706(2)(A), (C); see also CSX Transp., Inc. v.
Surface Transp. Bd., 

, 1063 (D.C. Cir. 2014);
Ass’n of Privacy Sector Colls. & Univs. v. Duncan, 

, 441 (D.C. Cir. 2012).

                             IV.

     For all three challenges, the first step of the Chevron
analysis proceeds in the same way. “We begin our analysis, as
always, with the statutory text.” Tesoro Alaska Co. v. FERC,

, 1038 (D.C. Cir. 2015). Congress having
expressly assigned the NCUA the power to define the
challenged terms, see 12 U.S.C. § 1759(g)(1), we may proceed
to Chevron’s second prong without further analysis, see U.S.
Telecom Ass’n v. FCC, 

, 717 (D.C. Cir. 2016);
Rush Univ. Med. Ctr. v. Burwell, 

, 760 (7th Cir.
2014); Comm’r v. Pepsi-Cola Niagara Bottling Corp., 

, 393 (2d Cir. 1968) (Friendly, J.) (“When Congress
has used a general term and has empowered an administrator
                               16
to define it, the courts must respect his construction if this is
within the range of reason.”).

     An express delegation of definitional power “necessarily
suggests that Congress did not intend the [terms] to be applied
in [their] plain meaning sense,” Women Involved in Farm
Econ. v. U.S. Dep’t of Agric., 

, 1000 (D.C. Cir.
1989), that they are not “self-defining,” 

the agency
“enjoy[s] broad discretion” in how to define them, Lindeen v.
SEC, 

, 653 (D.C. Cir. 2016). In Chevron terms,
Congress through explicit language “has directly spoken to the
precise question” of whether the identified terms must carry
certain 

. The answer is no. See
Buongiorno v. Sullivan, 

, 509 (D.C. Cir. 1990)
(Thomas, J.) (“When Congress expressly delegates the
authority to fill a gap in a statute, Congress speaks, in effect,
directly, and says, succinctly, that it wants the agency to
annotate its words.”). To hold otherwise at the first Chevron
step would “undermine” the ability of Congress to delegate
definitional power. Rush Univ. Med. 

.

    Consequently, we turn to whether the NCUA’s definitions
are “based on a permissible construction of the statute.”

.

                               V.

     Agency interpretations promulgated to fill an explicit
legislative gap “are given controlling weight unless they are
arbitrary, capricious, or manifestly contrary to the statute.”

; see also Pharm. Research & Mfrs.
of Am. v. FTC, 

, 204 (D.C. Cir. 2015).

    Under arbitrary and capricious review, “we may not
substitute our own judgment for that” of the agency. FERC v.
                               17
Elec. Power Supply Ass’n, 

, 782 (2016); accord
Dep’t of Commerce v. New York, 

, 2569 (2019).
Still, we are “not a ‘rubber stamp,’” 

;
“the agency must examine the relevant data and articulate a
satisfactory explanation for its action,” Motor Vehicle Mfrs.
Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co. (State
Farm), 

, 43 (1983). A rule is arbitrary and
capricious if (1) the agency “has relied on factors which
Congress has not intended it to consider”; (2) the agency
“entirely failed to consider an important aspect of the
problem”; (3) the agency’s explanation “runs counter to the
evidence before the agency”; or (4) the explanation “is so
implausible that it could not be ascribed to a difference in view
or the product of agency expertise.” 

we assess three definitional changes in the 2016
rule: (1) qualifying Combined Statistical Areas as local
communities; (2) eliminating the core requirement for local
communities based on Core Based Statistical Areas; and
(3) raising the population cap for rural districts. We sustain the
first and third amendments in full. As for the second, we hold
that it is rationally related to the Act’s text and purposes, but
that it is insufficiently explained.

                               A.

    The District Court rejected the first change because it
approved certain Combined Statistical Areas “no matter how
geographically dispersed and unconnected” the “members may
be.” Am. Bankers 

. To the court,
the approval did not fit with the term “local community,”
which, in its view, “encompass[es] an area no larger than a
county.” 

61. We respectfully disagree.
                               18
    The NCUA possesses vast discretion to define terms
because Congress expressly has given it such power. But the
authority is not boundless. The agency must craft a reasonable
definition consistent with the Act’s text and purposes; that is
central to the review we apply at Chevron’s second step. Here,
the NCUA’s definition meets the standard.

     We first focus on the text. Congress introduced the phrase
“local community” in the 1998 amendments. The word
“community” had a broad scope at the time. It meant not only
“society at large” but also a “body of individuals organized into
a unit or manifesting usu[ally] with awareness of some
unifying trait.”       See Community, WEBSTER’S THIRD
INTERNATIONAL DICTIONARY OF THE ENGLISH LANGUAGE
UNABRIDGED 460 (1993). The group could be “united by
historical consciousness or . . . common social, economic, and
political interests.” 

unifying trait could also be
simply “living in a particular place or region.” 

has recognized that the modifier “local”
“reflects congressional intent that it takes ‘a more circumspect
and restricted approach to chartering community credit
unions.’” Am. Bankers 

(citation
omitted); see also Oral Arg. Recording 3:16-18. Indeed,
Congress made clear its intention to “modif[y]” the “current
law regarding community credit unions” by adding the word
“local” to community. S. REP. NO. 105-193, at 6-7.

     Insertion of the modifier “local” before “community”
implies that the community “relate[s] to” a “particular limited
district” or is “confined to a particular place.” Local,
WEBSTER’S THIRD INTERNATIONAL DICTIONARY OF THE
ENGLISH LANGUAGE UNABRIDGED 1327 (1993). But that place
need not be the size of a county, as the District Court held. The
Supreme Court has recognized that the geographical areas
                               19
“need not be small.” First Nat’l Bank 

. And
if Congress wanted a local community to correspond to a
particular geographical unit, such as a county, “‘it easily could
have written’ that limitation explicitly.” NCUA Br. 25-26
(quoting Ali v. Fed. Bureau of Prisons, 

, 227
(2008)); see also Credit Union Nat’l Ass’n Amicus Br. 15. The
NCUA sensibly reads the term “local” to mean simply that the
community, regardless of shape or size, should be neither
“broad” nor “general.” Local, SUPRA, at 1327.

    To be clear, we do not hold today that the NCUA must
consider only bonds and social connections as understood in
1998. The parties agree, and thus we assume, that the NCUA,
despite its expressly delegated authority, must adopt a
definition consistent with what the term “local community”
meant in 1998, the time of its adoption.

     After consulting state statutes and invoking the canon of
noscitur a sociis, the District Court developed a rather size-
restrictive meaning for the phrase. See Am. Bankers 

. But we do not think that defining a “local
community” to refer to an area larger than a county is an
unreasonable interpretation of the Act’s text.

     We receive little guidance from the state statutes in effect
in 1998. Indeed, some of the statutes considered “local
communities” to be quite large. See, e.g., ALASKA STAT.
§ 18.66.990(7) (1998) (defining “local community entity” as “a
city or borough or other political subdivision of the state, a
nonprofit organization, or a combination of these” (emphasis
added)); see also NCUA Br. 25 n.4.

    We also reject the District Court’s invocation of the
noscitur a sociis canon. When several terms “are associated in
a context suggesting that [they] have something in common,
                                20
they should be assigned a permissible meaning that makes
them similar.” ANTONIN SCALIA & BRYAN A. GARNER,
READING LAW: THE INTERPRETATION OF LEGAL TEXTS 195
(2012). But the “substantive connection, or fit, between” the
terms here – local community, neighborhood, and rural
district – is “not so tight or so self-evident.” Graham Cty. Soil
& Water Conservation Dist. v. United States ex rel. Wilson, 

, 288 (2010). Only the word “neighborhood” has a
dictionary definition clearly suggesting a region of small size;
the phrase “local community” does not, as we have explained
above, and neither does “rural district,” as we explain below.
A size restriction would impermissibly “submerge[]” the
independent “character” of the latter two terms. Babbitt v.
Sweet Home Chapter of Cmtys. for a Great Or. (Sweet Home),

, 702 (1995) (citation omitted).

     The Association also points to other textual indicators.
But contrary to what it suggests, see Am. Bankers Ass’n Br.
27-28, we see nothing in the record suggesting that “local
community” is a term of art. The Association also says the
usage of word “local” in two other federal statutes indicates
that rules permitting coverage areas of larger than a county
would be manifestly contrary to the Act. See 

see also
15 U.S.C. § 2203 (1998) (defining “local” as “city, town,
county, . . . or other political subdivision of a State”); 18 U.S.C.
§ 666 (1998) (same). (The Association pointed to other laws,
but they did not exist in 1998. See National Defense
Authorization Act for Fiscal Year 2010, Pub. L. No. 111-84,
§ 4703(b), 123 Stat. 2190, 2837 (2009); PRIME Act Grants, 66
Fed. Reg. 29,010 (May 29, 2001).) True enough. But the two
statutes address materially distinct issues – fire prevention and
embezzlement – and thus do not indicate that the term “local”
must imply a size limitation here. See Envtl. Def. v. Duke
Energy Corp., 

, 574 (2007) (“[M]ost words have
                                21
different shades of meaning and consequently may be variously
construed . . . in different statutes . . . .” (citation omitted)).

     In addition to being consistent with the Act’s text, the
Combined Statistical Area definition rationally advances the
Act’s underlying purposes. In the 1998 amendments, Congress
made two relevant findings about purpose. First, legislators
found “essential” to the credit-union system a “meaningful
affinity and bond among members, manifested by a
commonality of routine interaction[;] shared and related work
experiences, interests, or activities[;] or the maintenance of an
otherwise well-understood sense of cohesion or identity.” § 2,
112 Stat. at 914. Second, Congress highlighted the importance
of “credit union safety and soundness,” because a credit union
on firm financial footing “will enhance the public benefit that
citizens receive.” 

history also confirms the
importance of common bonds, see S. REP. NO. 73-555, at 2; see
also H.R. REP. NO. 105-472, at 12; H.R. REP. NO. 73-2021, at
2; 144 CONG. REC. S9094 (daily ed. July 28, 1998) (statement
of Sen. Mikulski); 

(daily ed. July 24, 1998)
(statement of Sen. Moseley-Braun), and economic “integrity,”
S. REP. NO. 105-193, at 3; see also 144 CONG. REC. S9094
(statement of Sen. Mikulski); 

(statement of Sen.
Moseley-Braun), to federal credit unions.

     We recognize that there may be some tension between the
Act’s principal purposes: A credit union with exceedingly
close ties among its members is unlikely to have a large enough
customer base to thrive economically. To the extent that such
tension exists, the Act leaves to the NCUA to strike a
reasonable balance. Congress was well aware that a viable
credit union might serve a relatively large geographical area.
See, e.g., H.R. REP. NO. 73-2021, at 2 (“[T]here are cases in
which communities and organizations cross State lines.”).
                              22
     The NCUA did just that in promulgating the Combined
Statistical Area definition. That definition allows for larger
community credit unions; the decision is consistent with
decades of history promoting the economic viability of credit
unions in the face of banks and other competing financial
institutions. Nonetheless, the NCUA struck a balance by
ensuring that members within the local community maintain
somewhat of a commuter relationship with each other. As the
Association even acknowledges, commuting patterns “may
sometimes serve as a proxy for community interaction.” Am.
Bankers Ass’n Br. 38 n.25. We see nothing irrational about
adopting the factor as a proxy for the common bond
contemplated by Congress. Perhaps we would have made a
different call had we been the policymakers. Perhaps we would
have sought a tighter bond. Or perhaps we would not have
prioritized credit-union growth. See Iowa Bankers Ass’n
Amicus Br. 24-25. But we must “refrain from substituting
[our] own interstitial lawmaking for that” of the agency.
Household Credit Servs., Inc. v. Pfennig, 

, 244
(2004) (citation omitted).

     The NCUA also reasonably explained its amendment to
the “local community” definition. To begin with, the agency
reasonably circumscribed the size of a local community under
the Combined Statistical Area rule by imposing a 2.5 million-
person population limit. Used by the OMB in analogous
circumstances, the cap is a “logical breaking point in terms of
community cohesiveness with respect to a multijurisdictional
area.” Chartering and Field of Membership for Federal Credit
Unions, 75 Fed. Reg. 36,257, 36,259 (June 25, 2010).
Moreover, the NCUA reasonably relied on its prior experience
with Core Based Statistical Areas, which no party disputes can
serve as local communities under the Act. The agency noted
that the average geographic size of Combined Statistical Areas
with populations of up to 2.5 million people is 4,553 square
                                23
miles, and that the average size of the Core Based Statistical
Areas approved by the NCUA as local communities is 4,572
square miles. See 81 Fed. Reg. at 88,414-15. Given that
similarity in size, the NCUA reasonably considered adopted its
population-limited Combined Statistical Areas standard.

     Finally, the NCUA’s definition does not readily create
general, widely dispersed regions. Cf. First Nat’l Bank 

(indicating that community credit unions may not
be “composed of members from an unlimited number of
unrelated geographical units”). Combined Statistical Areas are
geographical units well-accepted within the government. See
81 Fed. Reg. at 88,414. Because they essentially are regional
hubs, the Combined Statistical Areas concentrate around
central locations. The OMB carved out the areas so that their
constituent parts share commuting ties and they reflect
“broader social and economic interactions.” OFFICE OF MGMT.
& BUDGET, SUPRA, at app. 2-3. The NCUA rationally believed
that such “real-world interconnections” would qualify as the
type of mutual bonds suggested by the term “local
community.” Credit Union Nat’l Ass’n Amicus Br. 9. Thus,
the agency reasonably determined that Combined Statistical
Areas “simply unif[y], as a single community,” already
connected neighboring regions. See 81 Fed. Reg. at 88,415.

     The Association raises a potpourri of objections to the
NCUA’s decision-making. See Am. Bankers Ass’n Br. 33-48.
Virtually all of its gripes are forfeited because it failed first to
raise them to the agency, see, e.g., Koretoff v. Vilsack, 

, 397-98 (D.C. Cir. 2013) (per curiam), or lack merit
because they involve outdated “local community” definitions,
which either did not allow for the qualification of certain
geographical areas as local communities or lacked a population
cap of 2.5 million people.
                              24
     But one of the complaints is “deserving of sustained
consideration.” Ortiz v. United States, 

, 2173
(2018). The Association contends that some large Combined
Statistical Areas are so sprawling that the NCUA’s definition,
which treats a 2.5-million-person portion of them as a local
community, must be unreasonable. Recall that a Combined
Statistical Area is a regional hub with more than one urban
center. Under the OMB’s technical specifications, each center
need not be connected to every other by commuting ties; rather,
each need only have ties to one other center within the hub. As
the Association colorfully puts it, the OMB may designate as a
Combined Statistical Area a mere “daisy chain” of urban
centers “that are linked to their neighbors but have nothing to
with those at the other end of the chain.” See Am. Bankers
Ass’n Br. 36 (citation omitted). Theoretically, the Association
continues, certain 2.5-million-person communities might bring
together parts of different urban centers with no connection
with one another. The Association also suggests that the rule
might permit local communities comprising non-contiguous
portions of a Combined Statistical Area. See 

     We understand the Association’s argument to be attacking
the Combined Statistical Area definition as unreasonable. To
the Association, the NCUA failed sufficiently to consider the
potential for the rule to create unreasonable results. One
hypothetical application disturbed the District Court here: the
prospect that, within the Combined Statistical Area including
the District of Columbia and counties from three states
(Maryland, Pennsylvania, and Virginia), one could create a
local community bringing together Doylesburg, Pennsylvania,
and Partlow, Virginia, towns located 200 miles apart. See Am.
Bankers 

.

    We might well agree with the District Court that the
approval of such a geographical area would contravene the Act.
                               25
But even so, the Association would need much more to mount
its facial pre-enforcement challenge in this case. As the
Supreme Court repeatedly has held, “the fact that petitioner can
point to a hypothetical case in which the rule might lead to an
arbitrary result does not render the rule” facially invalid. Am.
Hosp. Ass’n v. NLRB, 

, 619 (1991); see also EPA
v. EME Homer City Generation, L.P. (EME Homer), 

, 524 (2014) (“The possibility that the rule, in uncommon
particular applications, might exceed [the agency]’s statutory
authority does not warrant judicial condemnation of the rule in
its entirety.”); INS v. Nat’l Ctr. for Immigrants’ Rights, Inc.,

, 188 (1991) (“That the regulation may be invalid
as applied in s[ome] cases . . . does not mean that the regulation
is facially invalid because it is without statutory authority.”);
cf. Barnhart v. Thomas, 

, 29 (2003) (“Virtually
every legal (or other) rule has imperfect applications in
particular circumstances.”).

     Here, the Association’s complaint and the District Court’s
accompanying worry strike us as too conjectural. The NCUA
must assess the “economic advisability of establishing” the
proposed credit union before approving it, 12 U.S.C. § 1754,
and as part of the assessment, the organizers must propose a
“realistic” business plan showing how the institution and its
branches would serve all members in the local community, see
12 C.F.R. pt. 701, app. B, ch. 1 § IV.D. The Association has
failed to demonstrate the plausibility of a local community that
is defined like the hypothetical narrow, multi-state strip and
accompanies a realistic business plan. And if the agency were
to receive and approve such an application, a petitioner can
make an as-applied challenge. See, e.g., EME 

; 

. The Association has
succeeded in such challenges in the past. See, e.g., Am.
Bankers Ass’n v. NCUA, No. 1:05-CV-2247, 

, at *1 (M.D. Pa. July 21, 2008); Am. Bankers Ass’n v.
                               26
NCUA, 

, 1074 (D. Utah 2004). Thus, we
reject the facial attack on the amended “local community”
definition involving Combined Statistical Areas.

                               B.

     We turn to the next rule change. The District Court upheld
the eliminated core requirement for a local community based
on a Core Based Statistical Area. The court acknowledged that
defining a local community without its urban core “does not
alter the . . . common bond” shared by the members in the
remainder. Am. Bankers 

.
Meanwhile, the court accepted the agency’s response to the
claim that the new definition encouraged redlining – a broad
category of lending practices with negative impact on “areas
where low-income and minority populations are concentrated.”

(quoting 81 Fed. Reg. at 88,413).

     Like the District Court, we hold that the eliminated core
requirement is consistent with the Act’s text and purposes.
Still, we see merit in the Association’s redlining argument and
thus hold the definitional change to be arbitrary and capricious.

                               1.

     We will sustain the eliminated core requirement if it
reflects a reasonable interpretation of “local community” that
is rationally related to the dual purposes of promoting credit-
union growth and ensuring some cohesion among members. It
does. Omission of the urban core from proposed geographical
area will permit community credit unions to reach new
members in the suburban parts of the Core Based Statistical
Area and thus to maintain a healthy membership. Because the
suburbs under the OMB’s definition have substantial
commuting ties to the urban cluster, they all will be “within a
                              27
feasible commuting radius” and thus “share[] at least some
geographic ties.” Am. Bankers 

.
Those bonds do not disappear if the local community lacks the
core. The Association seeks greater “assurance[s]” of a
meaningful bond among members, Am. Bankers Ass’n Br. 66-
67; see also Am. Bankers Ass’n Reply 11, but we do not
replace the agency’s policy judgment with ours.

     We also do not believe that the eliminated core
requirement would create sprawling, boundless geographical
regions. No one disputes that, as a general matter, a
membership area comprising an intact Core Based Statistical
Area will satisfy any definition of “local community.” If the
local community with the core poses no problem, we fail to see
how a local community without one would. And even as to
Core Based Statistical Areas that do not qualify as local
communities (because they have populations of more than 2.5
million), the geographical ties ensure that the proposed
membership area will still be contained within the boundaries
of a single, well-recognized metropolitan region. A single
region is not what concerned the Supreme Court in First
National Bank III. 

.

     The Association objects to the expansive hypothetical
membership fields, highlighting two Core Based Statistical
Areas whose populations exceed the NCUA’s cap of 2.5
million. The Association asserts that the rule change “makes it
much easier to unite far-distant edges” of those sprawling areas
in a single membership area. Am. Bankers Ass’n Reply 9. Fair
point. But economic realities do not make it plausible that
organizers would propose such a local community or that the
NCUA would approve it. Like the Combined Statistical Area
definition, the eliminated core requirement does not become
facially infirm because of farfetched hypotheticals. To the
extent they occur in the future, troublesome rule applications
                              28
might be subject to as-applied challenges. See Am. Bankers

.

                               2

    Despite the eliminated core requirement’s consistency
with the Act, we cannot sustain the definitional change because
the NCUA has not adequately explained it.

     The Association contends that the rule change “effectively
allows credit unions to engage in ‘redlining’ by denying service
to urban areas with large numbers of minority and lower-
income residents.” Am. Bankers Ass’n Br. 65. The NCUA
attempted to address this concern in the preamble. At first
blush, the agency’s statements appear persuasive. Still, the
Association persuasively argues that the response fails to
“consider an important aspect” of the redlining issue or is
otherwise “so implausible” as to be unreasonable. State 

. Thus, the eliminated core requirement is
arbitrary and capricious.

     During the notice-and-comment proceedings, the
Association warned against redlining and objected that
community credit unions could now “serv[e] wealthier
suburban counties and exclud[e] markets containing low-
income and minority communities that reside in the core area.”
Letter from James Chessen, Exec. Vice President & Chief
Economist, Am. Bankers Ass’n, to Gerard S. Poliquin, Sec’y
of the Board, Nat’l Credit Union Admin. (Feb 5, 2016), Am.
Bankers Ass’n v. NCUA, No. 1:16-cv-02394-DLF (D.D.C.
filed Aug. 23, 2017), ECF No. 26-1 at 228. Fairly read, the
Association’s objection is not to traditional redlining, which
encompasses the refusal to make loans in low-income or
minority neighborhoods within a service area, see 

; see also 

, 355. Federal
                               29
credit unions “cannot ‘redline’” in the traditional sense because
“everyone in a credit union’s ‘community’ is a member who is
eligible to take advantage of credit union services” and credit
unions “by definition cannot reach beyond their member
communities to offer credit to the general public.” 

. But a community credit union can engage in
more unconventional redlining practices: “gerrymander[ing] to
create its own community of exclusively higher-income
members.” 

We think it evident that the Association
was focusing on such gerrymandering.

     In response, the NCUA acknowledged that it originally
kept the core requirement as a benefit to “low-income and
underserved populations.” 81 Fed. Reg. at 88,413. Some
commenters criticized the core requirement as failing to
achieve that goal; it caused community credit unions “to
sacrifice service to other areas” within the Core Based
Statistical Area, 

service arguably could benefit
poor and minority customers residing outside the core, see 

(remarking on the importance of credit unions
“providing financial services to low income and underserved
populations without regard to where they are located within a
community, i.e., beyond its ‘core area’”).

    Still, the agency dismissed the redlining concern on other
grounds, pointing to its “supervisory process to assess [credit-
union] management’s efforts to offer service to the entire
community [the credit union] seeks to serve.” 

focused on two aspects of its process. First, the agency touted
its “annual evaluation,” which “encompasses [the credit
union]’s implementation of its business and marketing plans.”

“[e]xperience” as support, the agency identified
the evaluation as a “more effective means” than a core
requirement to “ensur[e] that the low-income and underserved
populations are fairly served.” 

evaluations
                              30
confirm that the agency has had “success in providing financial
services to low-income and underserved populations without
regard to where they are located within” the membership area.

Second, the NCUA noted its “mandate to
consider member complaints alleging discriminatory practices
affecting low-income and underserved populations, such as
redlining, and to respond as necessary when such practices are
shown to exist.” 

of the NCUA’s supervisory process fail to
address the redlining issue raised by the Association. The
annual evaluation process might be an adequate response to
traditional redlining, because it might ensure that community
credit unions adequately serve poor and minority residents
living in their local communities. But we do not see how it
fixes gerrymandering or the potential discriminatory economic
impact on urban residents. See State 

(requiring the agency to consider relevant factors). The annual
evaluation process necessarily does not come into effect until
the NCUA already has approved the charter, business plan, and
proposed local community. See Iowa Bankers Ass’n Amicus
Br. 12. And nothing in the record suggests that business plans
may focus on residents residing outside the finalized
membership area; in fact, the law forbids federal credit unions
from serving those residents.

     The complaint process also does not work in the
gerrymandering context.        As the preamble points out,
complaints are raised by the membership, which would not
include the affected urban residents because of the rule change.
It seems quite implausible, absent some contrary evidence the
agency failed to detail, that members will file grievances based
on gerrymandering harms suffered by residents outside the
coverage area. See State 

(rejecting
implausible explanations).
                               31

     The NCUA attempts to defend the rule change by offering
a buffet of other potential rationales in its briefing and at oral
argument. They all fail because the agency did not adopt them
when promulgating the rule change. See State 

(noting that courts may not “supply a reasoned basis for
the agency’s action that the agency itself has not given”
(quoting SEC v. Chenery Corp., 

, 196 (1947))).
We pause, though, to caution the NCUA about two proffered
reasons. At oral argument, the government counsel suggested
that the agency may reject proposed local communities if it
suspects they discriminate against residents in the urban core.
See Oral Arg. Recording 5:28-49, 13:43-14:09. But current
reviewing guidelines do not indicate that the agency looks for
such discrimination. See 12 C.F.R. pt. 701, app. B, ch. 1
§ VII.A. And the NCUA made the opposite representations to
the District Court. See Transcript of Motion Hearing at 48:23-
49:7 (“[District Court]: . . . If a credit union comes to the
agency and says I want to serve X area, either in a rural district
or a combined statistical area, and they meet the definition, the
agency has no authority to reject that application, as long as the
credit union can demonstrate that they can serve the
area? [NCUA]: . . . I think that’s probably right, your
Honor.”), Am. Bankers Ass’n v. NCUA, No. 1:16-cv-02394-
DLF (D.D.C. filed Mar. 27, 2018), ECF No. 33. The
government counsel also suggested that community credit
unions already cover the vast majority of urban cores. See Oral
Arg. Recording 8:10-22, 12:30-41; see also 

cf.
NCUA Reply 27-28. Perhaps, but the current record does not
support the assertion.
                               32
                               C.

     Finally, the District Court rejected the increased
population cap for rural districts. The court worried that a rural
district satisfying the new, higher limit could be too large or
could contain “numerous urban centers.” Am. Bankers 

. Because the rule qualified certain
districts “no matter [their] geographic size or the number or
size of metropolitan areas falling within [their] proposed
boundaries,” the court held that the NCUA’s new definition is
unreasonable. 

But in our view, the new “rural
district” definition is reasonable.

     As suggested above, we assume that the NCUA must
adopt definitions consistent with the statutory terms as
understood in 1934, not today. The terms “rural” and “district”
do not connote specific population or geographical constraints.
See First Nat’l Bank 

(noting that markets
“need not be small”). A “district” means a “portion of a state,
county, country, town, or city . . . made for administrative,
electoral, or other purposes.” District, WEBSTER’S NEW
INTERNATIONAL DICTIONARY OF THE ENGLISH LANGUAGE 649
(2d ed. 1934) (emphasis added). The district may be “of
undefined extent.” 

indicates a
“country” or “agricultur[al]” lifestyle, as opposed to that of a
“city or town.” Rural, WEBSTER’S NEW INTERNATIONAL
DICTIONARY OF THE ENGLISH LANGUAGE 1861 (2d ed. 1934).
“Rural” lands generally “resembl[e]” the “country[side].” 

the 1-million-person cap prevents the rural
district from resembling the countryside. And one of the
unchallenged restrictions helps provide a rural character to
such districts. Either most residents in the proposed district
live on rural land, or the population density is 100 or fewer
people per square mile. See 81 Fed. Reg. at 88,440.
                               33
Accordingly, even if most residents live on urban areas in the
rural district, those areas will be surrounded by rural land.
That’s because 100 people per square mile – or one person for
every six or so acres – is a rural population density.

     As for size, the contemporaneous definition of “district”
reassures us that rural districts may cross state lines. And a
second unchallenged restriction assures that the 1-million-
person cap will not support gigantic or straggly rural districts.
As the rule explains, the boundaries of a community credit
union’s district may not “exceed the outer boundaries of the
states” immediately surrounding the state where the proposed
credit union would have its headquarters. 81 Fed. Reg. at
88,440. Thus, even though the population density of the entire
United States is less than 100 people per square mile, see
NCUA Br. 56 n.44; Oral Arg. Recording 39:34-38, the
geographical limitation forces a rural district to be much
smaller. We are confident that such districts will not be so
enormous as to amount to federations of “unrelated geographic
units.” See First Nat’l Bank 

.

     In arguing that the amended “rural district” definition is
unreasonable, the Association relies heavily on the interpretive
analysis performed by the District Court. See Am. Bankers

. The court’s core contention
is that we must construe “rural” and “district” together as a
specialized phrase that meant, in 1934, an “area[] much smaller
than a state.” 

For support, the District Court not only
consulted two 1934-era dictionaries, results of a Westlaw
search of contemporaneous opinions, and a database containing
historical uses of the phrase, but also invoked the noscitur a
sociis canon. See 

    The proffered definitional evidence is pretty thin. The
1934-era dictionaries described what the specialized phrase
                              34
“rural district” meant in Great Britain, not in the much larger
and more expansive United States, which by the 1930s
encompassed forty-eight continental states. See, e.g., Rural
District, WEBSTER’S NEW INTERNATIONAL DICTIONARY OF THE
ENGLISH LANGUAGE 1861 (1934) (“in England, a subdivision
of an administrative county embracing . . . county parishes”).

     The proffered Westlaw search results included scores of
federal and state judicial and administrative opinions. See
Plaintiff’s Supplemental Memorandum app. 6, Am. Bankers
Ass’n v. NCUA, No. 1:16-cv-02394-DLF (D.D.C. filed Mar.
16, 2018), ECF No. 32-6. But we find little of use in those
documents. Many of those opinions do not discuss size or
population. See, e.g., Nicolai v. Wis. Power & Light Co., 

, 282-83 (Wis. 1936). Those that do involve specific
rural districts whose sizes were between a town and a state.
See, e.g., Sarther Grocery Co. v Comm’r, 

, 1274
(1931). But none declares that rural districts by definition are
restricted to any particular size or population.

    The District Court next turned to the Corpus of Historical
American Usage, a free and public online database. See Am.
Bankers 

& n.5. The database notes
197 mentions of the phrase in the first half of the twentieth
century and only 37 in the second. See Search Results for
RURAL DISTRICTS and RURAL DISTRICT, CORPUS OF
HISTORICAL AMERICAN ENGLISH, https://www.english-
corpora.org/coha/ (follow “List” hyperlink; then search
matching strings field for “RURAL DISTRICT”). The
perceived drop-off led the District Court to determine that,
“even if rural district does not carry meaning distinct from its
individual words today, it did in 1934.” Am. Bankers 

.
                               35
     Although federal courts may use “crude[]” searches on
databases to learn of ambiguities in a statutory term,
Muscarello v. United States, 

, 129-30 (1998), the
search here did not suffice to show that the agency’s definition
was unreasonable. Much more is required to cabin the
agency’s discretion. A search of a commercial database,
ProQuest, reveals that the phrase appeared at least 500 times in
the second half of the century. We are not confident that the
proffered evidence establishes a particular historical trend.

     The District Court lastly invoked noscitur a sociis. See
Am. Bankers 

-69. The canon is
inapposite. Recall that the term “rural district” was listed with
“community,” not “local community,” in 1934. At the time,
“community” could refer to “[s]ociety at large” or a
“commonwealth or state.” Community, WEBSTER’S NEW
INTERNATIONAL DICTIONARY OF THE ENGLISH LANGUAGE 452
(1934). Indeed, we said that the word is “too indefinite to be
used for purposes of exact measurement in terms of acres or
square miles.” Lukens Steel Co. v. Perkins, 

, 631
(D.C. Cir. 1939), rev’d sub nom., Perkins v. Lukens Steel Co.,

(1940). Thus, we do not see size or population
as a connection between the terms.

     The increased population cap is consistent with not only
the Act’s text but also its purposes. For instance, as the
preamble noted, the expanded definition allows community
credit unions in rural districts to reach more “persons of modest
means who may reside” in those areas. 81 Fed. Reg. at 88,416;
see also § 2(1), (2), 112 Stat. at 913.

     And the change is neither arbitrary nor capricious. The
new rule explains that it provides a “level of operating
efficiencies and scale” making rural districts attractive options
for prospective credit unions. 81 Fed. Reg. at 88,416. “[A]
                              36
sufficiently large population base is essential to enable” them
“to offer financial services economically.” 

The
NCUA also chose the 1-million figure based on prior
experience. The agency noted that it had approved of eight
districts with populations of more than 250,000, and that one
of them already had exceeded 1 million. See 

a
1 million precedent in one state,” the agency felt justified in
having a “fixed 1 million population cap for the other 49
states.” 

We cannot say this policy
choice lacks rational explanation.

     The Association says the NCUA failed to consider prior
agency decisions. See Am. Bankers Ass’n Br. 61. But we see
no discrepancy and thus summarily reject the objection. The
Association also turns to troubling hypothetical examples of
rural districts with unruly shapes and those with dense urban
areas such as Denver, Colorado. See NCUA Br. 55-61; Oral
Arg. Recording 37:55-38:05. Again, such implausible outliers
do not impugn the rule’s general reasonableness.

                              VI.

     We now consider the remedy. To sum up, we hold that
defining certain Combined Statistical Areas or portions of them
as local communities and raising the population cap for rural
districts are consistent with the Act and reasonably explained.
Thus, we sustain both aspects of the challenged rule. We also
leave undisturbed the portion of the District Court’s opinion
that the parties do not contest on appeal.

     But we deem the eliminated core requirement to be
arbitrary and capricious. When a rule is contrary to law, the
“ordinary practice is to vacate” it. United Steel v. Mine Safety
& Health Admin., 

, 1287 (D.C. Cir. 2019); see
also 5 U.S.C. § 706(2) (noting that the “reviewing court
                                  37
shall . . . set aside” unlawful agency action). But in “rare
cases,” we will opt instead to remand without vacating the rule,
so that the agency can “correct its errors.” United 

; see also 5 U.S.C. § 702 (“Nothing
herein . . . affects . . . the power or duty of the court to . . . deny
relief on any . . . appropriate . . . equitable ground . . . .”); Oglala
Sioux Tribe v. U.S. Nuclear Regulatory Comm’n, 

,
536 (D.C. Cir. 2018). In considering whether to adopt the latter
equitable remedy, we balance “(1) the seriousness of the
deficiencies of the action, that is, how likely it is the agency
will be able to justify its decision on remand; and (2) the
disruptive consequences of vacatur.” United 

(citation omitted). A strong showing of one factor may
obviate the need to find a similar showing of the other. See,
e.g., Fox Television Stations, Inc. v. FCC, 

, 1049
(D.C. Cir. 2002) (holding that, because the agency likely could
justify its action on remand, the potential for disruption was
“only barely relevant”); Allied-Signal, Inc. v. U.S. Nuclear
Regulatory Comm’n, 

, 152 (D.C. Cir. 1993)
(determining that, because vacatur would give regulated parties
a “peculiar windfall,” the meager chance of justifying the
action was given “little weight” in the remedial analysis).

     The NCUA has not requested remand without vacatur in
this case. But because we have a “duty” to ensure the propriety
of the APA remedy, 5 U.S.C. § 702, we hold that we have the
discretion to raise the issue sua sponte, cf. Igonia v. Califano,

, 1387 (D.C. Cir. 1977); Gaddis v. Dixie Realty
Co., 

, 247 (D.C. Cir. 1969) (per curiam). Remand
without vacatur is appropriate here.

     We conclude that the NCUA might be able to offer a
satisfactory reason on remand. And as for disruptive effect, we
perceive a substantial likelihood that vacating the rule could
make it more difficult for some poor and minority suburban
                               38
residents to receive adequate financial services.           Even
temporarily depriving these members of the opportunity is
inequitable, because it would “set back” the Act’s objective of
offering financial services to people of small means. See Nat’l
Res. Def. Council v. EPA, 

, 1374 (D.C. Cir.
2007); see also Advocates for Highway & Auto Safety v. Fed.
Motor Carrier Safety Admin., 

, 1152 (D.C. Cir.
2005) (declining to vacate rule because, in the interim, it “may
do some good, if it does anything at all”).

     Given the potential for sufficient justification and the
substantial likelihood of disruptive effect, we have a rare case
in which vacatur is inappropriate. See United 

. Thus, we remand without vacating the relevant part
of the 2016 rule. We trust that the agency will act
expeditiously. See EME Homer City Generation, L.P. v. EPA,

, 132 (D.C. Cir. 2015).

                           *   *    *

     In short, we reverse the challenged portions of the District
Court’s summary judgment order. With respect to the
qualification of certain Combined Statistical Areas as local
communities and the increased population cap for rural
districts, we direct the District Court to issue summary
judgment in favor of the NCUA. As to the elimination of the
urban-core requirement for local communities based on Core
Based Statistical Areas, we direct the District Court to issue
summary judgment in favor of the Association and to remand,
without vacating, the relevant portion of the 2016 rule for
further explanation.

                                                    So ordered.