The opinion
FOR PUBLICATION
UNITED STATES COURT OF APPEALS
FOR THE NINTH CIRCUIT
JEFFREY POWERS; DEAVIN No. 24-6576
SESSOM; LAURIEANN WRIGHT;
D.C. No.
JOSEPH FIELDS; LAVON
2:22-cv-08357-
JOHNSON; JOSHUA ROBERT
DOC-KS
PETITT; NATIONAL VETERANS
Central District of
FOUNDATION,
California,
Los Angeles
Plaintiffs - Appellees,
ORDER
and
SAMUEL CASTELLANOS,
SHARDAY ANYADIEGWU,
BILLY EDWARDS, JESSICA
MILES, GLENN SURRETTE,
NARYAN STIBBIE, DOES 1-10,
Plaintiffs,
BRIDGELAND RESOURCES, LLC,
Intervenor-Plaintiff - Appellee,
v.
DENIS RICHARD MCDONOUGH;
STEVEN BRAVERMAN; KEITH
HARRIS; MARCIA L. FUDGE,
2 POWERS V. MCDONOUGH
Defendants - Appellants,
and
DOUGLAS GUTHRIE,
Defendant - Appellee,
----------------------------------------
BRENTWOOD SCHOOL,
Intervenor.
Filed September 11, 2026
Before: Consuelo M. Callahan, Roopali H. Desai, and Ana
de Alba, Circuit Judges;
Order;
Dissent by Judge Collins;
Dissent by Judge Bumatay.
POWERS V. MCDONOUGH 3
SUMMARY*
Veterans Affairs
The panel denied rehearing en banc in a case in which
the panel affirmed in part, reversed in part, vacated in part,
and remanded to the district court to enter judgment in a class
action brought by unhoused veterans with severe disabilities
and mental illnesses seeking to return the West Los Angeles
Veterans Affairs grounds to its intended use of housing
disabled veterans.
Dissenting from the denial of rehearing en banc, Judge
Collins, joined by Judges Bress, Bumatay, and Tung, wrote
that the panel’s decision, which upheld the district court’s
judgment and injunction requiring the Department of
Veterans Affairs to construct housing for the class members
based on § 504 of the Rehabilitation Act, disregards the
limitations that the Supreme Court has placed on disability
discrimination claims asserted under § 504 of the
Rehabilitation Act. The district court’s decision was based
on a deeply flawed legal foundation, and on an equally
flawed certification of a class action.
Dissenting from the denial of rehearing en banc, Judge
Bumatay, joined by Judges VanDyke and Tung, wrote that
while Judge Collins provided a persuasive explanation as to
why Plaintiffs-Appellees’ claims should fail on the merits,
the panel’s errors were even more fundamental. First, in
affirming the district court’s expansive injunction, the panel
wrongly took as given that the Rehabilitation Act includes
*
This summary constitutes no part of the opinion of the court. It has
been prepared by court staff for the convenience of the reader.
4 POWERS V. MCDONOUGH
an implied right of action. Second, the panel reduced the
class-action commonality requirement under Fed. R. Civ. P.
23(a)(2) to a mere pleading exercise.
ORDER
The panel unanimously voted to deny Appellants’
petition for rehearing en banc. The full court was advised of
the petition for rehearing en banc. A judge requested a vote
on whether to rehear the matter en banc. The matter failed
to receive a majority of the votes of the nonrecused active
judges in favor of en banc consideration. Fed. R. App. P. 40.
Judges Gould and Rawlinson did not participate in the
deliberations or vote in this case.
The petition for rehearing en banc (Dkt. No. 127) is
DENIED. Judge Collins’s and Judge Bumatay’s dissents
from the denial of rehearing en banc are filed concurrently
herewith. No further petitions for rehearing will be
entertained.
COLLINS, Circuit Judge, joined by BRESS, BUMATAY,
and TUNG, Circuit Judges, dissenting from the denial of
rehearing en banc:
The panel’s decision in this case confirms that
sympathetic facts make bad law. By all accounts, the
Department of Veterans Affairs (“VA”) has not done a good
job of providing for veterans in Los Angeles, particularly
homeless veterans. Its failures have drawn the scrutiny of
Congress, which has twice passed (in 2016 and 2021)
legislation specifically concerning homeless veterans in Los
Angeles, as well as the attention of the President, who last
POWERS V. MCDONOUGH 5
year issued an Executive Order that likewise focuses on the
VA’s west Los Angeles campus. But the answer to this
complex problem does not lie in the remarkable decree
upheld by the panel here.
That decree is based on the extraordinary theory that the
VA’s administration of its statutory program for providing
medical care to veterans unlawfully discriminates against
disabled veterans in violation of § 504 of the Rehabilitation
Act, 29 U.S.C. § 794, unless the VA also affirmatively
provides housing for those disabled veterans who are
homeless. See Powers v. McDonough, 163 F.4th 1162,
1186–88, 1194–96 (9th Cir. 2025). This attempt to use the
Rehabilitation Act to judicially engineer a dramatic
expansion in the scope of a federal benefits program violates
well-settled Supreme Court and Ninth Circuit precedent.
Based on that deeply flawed legal foundation, and on an
equally flawed certification of a class action, the district
court has decreed that the VA must undertake a judicially
micromanaged program of housing construction at the VA’s
west Los Angeles campus that (for now) will include 1,800
permanent and 750 temporary housing units for the
homeless, at a cost well exceeding $100 million. Id. at 1195.
Perhaps such a program should be adopted and funded by
Congress to address the suffering of homeless veterans in
southern California. But such judgments do not belong to
the courts, and disregard of settled law, even in the service
of worthy aims, should not be countenanced. We should
have reheard this case en banc, and I dissent from our failure
to do so.
I
In its current posture, this case involves a certified class
action pursued by six individual plaintiffs and one
6 POWERS V. MCDONOUGH
organizational plaintiff (the National Veterans Foundation)
on behalf of “[a]ll homeless veterans with Serious Mental
Illness [SMI] or Traumatic Brain Injuries [TBI], who reside
in Los Angeles County.” Plaintiffs named various
defendants, but the only relevant defendants under the
panel’s decision are three VA officials who are sued in their
official capacities (hereinafter, collectively, the “VA”). 1
Plaintiffs’ core allegation is that, in administering its
statutory obligation to “provide a complete medical and
hospital service for the medical care and treatment of
veterans,” 38 U.S.C. § 7301(b), the VA “do[es] not provide
adequate Permanent Supportive Housing to ensure that . . .
[Plaintiffs can] access the medical treatment and other
services for which they are eligible.” Specifically, Plaintiffs
contend that, in order to avoid discriminating against
disabled veterans in the provision of VA medical care in
violation of § 504 of the Rehabilitation Act,2 the VA must
1
The district court’s judgment had also applied to the Secretary of
Housing and Urban Development (“HUD”), but the panel held that, with
respect to the relevant claims that remain at issue, there was no legal
basis for extending the judgment against the VA to HUD. See Powers,
163 F.4th at 1188. The panel therefore directed that the judgment against
HUD be vacated in its entirety. Id. I do not question the correctness of
that holding.
2
As relevant here, § 504 of the Rehabilitation Act provides:
No otherwise qualified individual with a disability
in the United States, . . . shall, solely by reason of her
or his disability, be excluded from the participation in,
be denied the benefits of, or be subjected to
discrimination under any program or activity receiving
Federal financial assistance or under any program or
activity conducted by any Executive agency or by the
United States Postal Service.
POWERS V. MCDONOUGH 7
construct both permanent and temporary housing for
veterans on the campus that contains the VA’s West Los
Angeles Medical Center (respectively, the “Campus” and the
“Center”).
Plaintiffs asserted two distinct claims under § 504 of the
Rehabilitation Act, and these two claims provide the sole
basis on which the panel upheld the district court’s housing-
construction injunction. First, in what the panel referred to
as Plaintiffs’ “meaningful access claim,” Plaintiffs alleged
(1) that the VA’s failure to affirmatively provide permanent
supportive housing on or near the Campus amounted to
unlawful discrimination against the disabled because it
deprives Plaintiffs of “meaningful access” to the medical
care to which they are entitled; and (2) that providing such
housing would be a “reasonable accommodation.” Powers,
163 F.4th at 1186, 1194–95.
Second, Plaintiffs alleged that, by failing to provide such
housing, the VA administers medical benefits “in a manner
that denies veterans the benefits of . . . services, programs,
or activities in the most integrated setting appropriate to their
needs”; that the VA’s “denial of appropriate integrated
services to Plaintiffs is solely because of their disabilities”;
and that “Plaintiffs are institutionalized or placed at risk of
institutionalization because of [the VA’s] discrimination.”
The panel referred to this as an “Olmstead” claim, Powers,
163 F.4th at 1187, referring to Olmstead v. L.C. ex rel.
Zimring, 527 U.S. 581 (1999).3
29 U.S.C. § 794(a).
3
In a third claim under § 504 of the Rehabilitation Act, Plaintiffs
challenged as discriminatory certain actions that the VA took with
respect to the income qualifications for receiving housing that was
8 POWERS V. MCDONOUGH
After certifying a class as described earlier, the district
court held a bench trial in August 2024. In its post-trial
decision, the district court ruled for Plaintiffs on all of their
§ 504 claims. As a remedy for Plaintiffs’ meaningful access
and Olmstead claims, the district court “enjoined [the VA]
from failing to provide Plaintiffs and the Class they represent
with appropriate permanent supportive housing on or near
the [Campus] so they can reasonably access the health care
benefits for which they are eligible in the most integrated
setting appropriate to their needs.” The district court ordered
the VA, “[w]ithin six (6) months of this Court’s order,” to
“develop a plan for the construction of an additional 1,800
units of permanent supportive housing . . . to address veteran
homelessness,” which “shall be built and operational within
six (6) years of this order.” At the time, there were 233
permanent housing units on the West Los Angeles VA
Grounds.
The panel upheld the district court’s grant of class
certification, and it affirmed the judgment and housing
injunction against the VA based solely on the meaningful
access and Olmstead claims. Powers, 163 F.4th at 1184–86,
1194–96. In holding that Plaintiffs had been
discriminatorily denied meaningful access to VA healthcare,
the panel relied on the district court’s finding that “Plaintiffs
. . . require access to on-site healthcare at the Campus; yet
without permanent supportive housing on or near the
Campus, Plaintiffs are unable to access such treatment.” Id.
at 1186. As to the Olmstead claim, the panel concluded that
otherwise being provided by private developers under VA authority.
Because this third claim does not undergird the district court’s housing-
construction injunction, I do not address whether the panel correctly
affirmed any independent relief that the district court ordered with
respect to that separate claim. See Powers, 163 F.4th at 1188, 1195–96.
POWERS V. MCDONOUGH 9
“the district court properly held that the VA violated the
Rehabilitation Act by not providing class members their
disability healthcare benefits” in the “most integrated setting
appropriate to their needs, placing unhoused veterans at risk
of institutionalization.” Id. at 1187 (simplified). In
upholding the certification of a class, the panel rejected the
VA’s argument that the merits of the meaningful access and
Olmstead claims “turn on each veteran’s individual
circumstances”; instead, the panel held, the commonality
requirement of Federal Rule of Civil Procedure 23 was
satisfied because Plaintiffs sought relief from a “system-
wide practice or policy.” Id. at 1185.
II
The panel’s decision upholding the district court’s
judgment and injunction requiring the VA to construct
housing for the class members rests solely on the theory that,
under § 504 of the Rehabilitation Act, the failure to
affirmatively supply such housing constitutes discrimination
against disabled persons in the provision of healthcare by the
VA. By construing § 504 to require this sort of
“fundamental alteration in the nature” of the VA’s healthcare
program, Alexander v. Choate, 469 U.S. 287, 300 (1985)
(citation omitted), the panel’s decision contravenes the
limitations the Supreme Court has established concerning
claims under § 504. The panel’s deeply flawed decision
warranted rehearing en banc.
A
Section 504 of the Rehabilitation Act provides, as
relevant here, that “[n]o otherwise qualified individual with
a disability in the United States . . . shall, solely by reason of
her or his disability, be excluded from the participation in,
be denied the benefits of, or be subjected to discrimination
10 POWERS V. MCDONOUGH
. . . under any program or activity conducted by any
Executive agency.” 29 U.S.C. § 794(a). As the Supreme
Court has held, “[a]n otherwise qualified person is one who
is able to meet all of a program’s requirements in spite of his
[disability].” Southeastern Cmty. Coll. v. Davis, 442 U.S.
397, 406 (1979). The Davis Court also acknowledged,
however, that an agency’s structuring of a program may
present practical obstacles to participation by otherwise
qualified disabled persons and that, in some situations, the
agency’s “refusal to modify” such a program to
accommodate such persons “might become unreasonable
and discriminatory.” Id. at 412–13 (emphasis added). The
general scope of § 504’s prohibition on discrimination, and
the circumstances in which § 504 might require such
accommodating modifications, were subsequently addressed
in the Supreme Court’s seminal decision in Choate, and I
therefore begin by reviewing that decision at some length
before addressing how the panel’s decision contravenes the
settled construction of § 504.
In Choate, disabled Medicaid recipients sued under the
Rehabilitation Act to enjoin a change to a state Medicaid
program that “reduc[ed] from 20 to 14 . . . the number of
inpatient hospital days per fiscal year that [state] Medicaid
would pay hospitals on behalf of a Medicaid recipient.” 469
U.S. at 289. The plaintiffs argued that the “proposed 14-day
limitation on inpatient coverage would have a discriminatory
effect on the handicapped,” principally because “[s]tatistical
evidence . . . indicated that . . . 27.4% of all handicapped
users of hospital services who received Medicaid required
more than 14 days of care, while only 7.8% of
nonhandicapped users required more than 14 days of
inpatient care.” Id. at 289–90. The Court unanimously
rejected this claim.
POWERS V. MCDONOUGH 11
The Court first addressed the question of “whether proof
of discriminatory animus is always required to establish a
violation of § 504 and its implementing regulations, or
whether federal law also reaches action by a recipient of
federal funding that discriminates against the handicapped
by effect rather than by design.” Choate, 469 U.S. at 292.
In considering these contrasting interpretations of § 504, the
Court noted that there were substantial difficulties with both
of them.
On the one hand, the Court noted that the authorities
cited by the defendants did “not support [their] blanket
proposition that federal law proscribes only intentional
discrimination against the handicapped.” Choate, 469 U.S.
at 294. As the Court explained, the defendants’ reading of
§ 504 was hard to square with what the Court understood to
be the objectives of that provision. The Court stated that
“[d]iscrimination against the handicapped was perceived by
Congress to be most often the product, not of invidious
animus, but rather of thoughtlessness and indifference—of
benign neglect.” Id. at 295.4 Moreover, the Court noted,
“much of the conduct that Congress sought to alter in passing
the Rehabilitation Act would be difficult if not impossible to
reach were the Act construed to proscribe only conduct
fueled by a discriminatory intent.” Id. at 296–97. In
particular, the Court noted that “elimination of architectural
barriers was one of the central aims of the Act, yet such
barriers were clearly not erected with the aim or intent of
4
Although the Court’s analysis on this score was largely based on
legislative history, rather than on any purpose declared in, or discernible
from, the statutory text, I am bound by “controlling Supreme Court
precedent telling us what the purpose of [the statute] is, even if that
precedent relies on legislative history.” Rojas v. FAA, 989 F.3d 666, 681
n.6 (9th Cir. 2021) (en banc) (Collins, J., concurring).
12 POWERS V. MCDONOUGH
excluding the handicapped.” Id. at 297 (simplified). The
same concerns were present, the Court explained, with
respect to “discrimination in access to public transportation,”
“the discriminatory effect of job qualification procedures,”
and discrimination through “the denial of special educational
assistance for handicapped children.” Id. (simplified). The
Court concluded that Congress’s intention to remedy such
discrimination “would ring hollow if the resulting legislation
could not rectify the harms resulting from action that
discriminated by effect as well as by design.” Id.
On the other hand, the Court was also “troubl[ed]” by the
competing argument that § 504 should be interpreted “to
reach all action disparately affecting the handicapped.”
Choate, 469 U.S. at 298. “Because the handicapped
typically are not similarly situated to the nonhandicapped,”
that position would “in essence require each recipient of
federal funds first to evaluate the effect on the handicapped
of every proposed action that might touch the interests of the
handicapped, and then to consider alternatives for achieving
the same objectives with less severe disadvantage.” Id. The
Court concluded that there was no statutory evidence that
“Congress intended § 504 to be a National Environmental
Policy Act for the handicapped, requiring the preparation of
‘Handicapped Impact Statements’ before any action was
taken by a grantee that affected the handicapped.” Id. at
298–99. The Court thus concluded that, “just as there is
reason to question whether Congress intended § 504 to reach
only intentional discrimination, there is similarly reason to
question whether Congress intended § 504 to embrace all
claims of disparate-impact discrimination.” Id. at 299.
The Court accordingly held that “[a]ny interpretation of
§ 504 must . . . be responsive to two powerful but
countervailing considerations—the need to give effect to the
POWERS V. MCDONOUGH 13
statutory objectives and the desire to keep § 504 within
manageable bounds.” Choate, 469 U.S. at 299. The Court
therefore explicitly “reject[ed] the boundless notion that all
disparate-impact showings constitute prima facie cases
under § 504.” Id. Although the Court had strongly
suggested that the converse position was also wrong, and
that § 504 was not limited merely to intentional
discrimination, the Court found it unnecessary to address
that issue. That was true, the Court stated, because even
“assum[ing] without deciding that § 504 reaches at least
some conduct that has an unjustifiable disparate impact upon
the handicapped,” the § 504 claim asserted in that case
failed. Id.
On the assumption that § 504 reached some disparate
impacts, the Court proceeded to consider “which disparate
impacts § 504 might make actionable.” Choate, 469 U.S. at
299. The “proper starting point,” the Court stated, was its
prior decision in Davis, where the plaintiff’s Rehabilitation
Act claim was properly rejected “because the further
modifications [the plaintiff] sought would have
compromised the essential nature of the [nursing] program”
to which she sought admittance. Id. at 300. This analysis in
Davis, the Choate Court explained, established the following
governing rule: “[W]hile a [defendant] need not be required
to make ‘fundamental’ or ‘substantial’ modifications to
accommodate the handicapped, it may be required to make
‘reasonable’ ones.” Id. Put another way, “an otherwise
qualified handicapped individual must be provided with
meaningful access to the benefit that the [defendant] offers,”
and “to assure meaningful access, reasonable
accommodations in the grantee’s program or benefit may
have to be made.” Id. at 301. But the Court reiterated that
the “reasonable accommodations” needed to provide
14 POWERS V. MCDONOUGH
“meaningful access” do not require fundamental
modifications to the program. Id. at 300–01.
Applying these principles, the Court concluded that
“[t]he 14-day limitation [did] not deny [the plaintiffs]
meaningful access to [state] Medicaid services or exclude
them from those services.” Choate, 469 U.S. at 302. As the
Court explained, “it cannot be argued that ‘meaningful
access’ to state Medicaid services will be denied by the 14-
day limitation on inpatient coverage,” because “[t]he
reduction in inpatient coverage will leave both handicapped
and nonhandicapped Medicaid users with identical and
effective hospital services fully available for their use, with
both classes of users subject to the same durational
limitation.” Id. The Court also regarded as “simply
unsound” the plaintiffs’ argument that “their greater need for
prolonged inpatient care means that, to provide meaningful
access to Medicaid services, [the State] must single out the
handicapped for more than 14 days of coverage.” Id. at 302–
03. As the Court put it, “[s]ection 504 does not require the
State to alter [the] definition of the benefit being offered
simply to meet the reality that the handicapped have greater
medical needs.” Id. at 303. At bottom, the plaintiffs sought
to “redefine” the program, which went “well beyond the
accommodations that are required under Davis.” Id. at 308.
Consistent with Choate, this court has recognized that a
“meaningful access” claim is simply one form of a
“reasonable accommodation” claim and that “[r]easonable
accommodation does not require an organization to make
fundamental or substantial alterations to its programs.”
Mark H. v. Hamamoto, 620 F.3d 1090, 1098 (9th Cir. 2010).
“[E]ntities are required only to make reasonable changes in
existing policies in order to accommodate individuals’
disabilities” and “are not required to create new programs
POWERS V. MCDONOUGH 15
that provide heretofore unprovided services to assist
disabled persons.” Townsend v. Quasim, 328 F.3d 511, 518
(9th Cir. 2003) (emphasis altered). “In no event is the entity
required to undertake measures that would impose an undue
financial or administrative burden or effect a fundamental
alteration in the nature of the service.” Where Do We Go
Berkeley v. California Dep’t of Transportation, 32 F.4th
852, 862 (9th Cir. 2022) (simplified) (quoting Tennessee v.
Lane, 541 U.S. 509, 532 (2004)).
B
Under these settled principles, the panel’s upholding of
Plaintiffs’ “meaningful access” claim and of the district
court’s resulting injunction squarely contravenes Supreme
Court and Ninth Circuit precedent.
1
As noted, Plaintiffs’ theory is that the absence of VA
housing in the vicinity of the VA Center amounts to a
discriminatory failure to accommodate their disabilities in
the VA’s provision of medical care, and their requested
remedy—which the district court granted—is for the VA to
undertake affirmative efforts to construct housing for them.
Powers, 163 F.4th at 1180. The panel concluded that this
unprecedented decree does not require the VA to
“fundamentally alter its programs,” but instead “simply
expands—not alters—the program.” Id. at 1195. This
holding is legally erroneous and contrary to precedent.
“Under the test outlined in Choate,” the court must “first
consider the nature of the benefit [the plaintiffs] were
allegedly denied.” Doe v. CVS Pharmacy, Inc., 982 F.3d
1204, 1210 (9th Cir. 2020) (emphasis added), cert.
dismissed, 142 S. Ct. 480 (2021). The benefit that Plaintiffs
16 POWERS V. MCDONOUGH
were allegedly discriminatorily denied here is the “medical
care and treatment” provided to them by the VA. 38 U.S.C.
§ 7301(b); see also 38 C.F.R. § 17.38(a). As the panel
correctly noted, see Powers, 163 F.4th at 1183, Plaintiffs
expressly concede that the “permanent supportive housing”
they seek is not among the “benefits” to which they are
entitled under the medical program at issue. As set forth in
the applicable regulations, the “medical benefits package” at
issue includes “[o]utpatient medical, surgical, and mental
healthcare,” “[i]npatient hospital, medical, surgical, and
mental healthcare,” “[p]rescription drugs,” and
“[e]mergency care in VA facilities,” and it does not include
permanent supportive housing. 38 C.F.R. § 17.38(a). 5
Rather, as the panel explained, Plaintiffs’ argument is that
“housing is a reasonable accommodation that is necessary
for them to access the benefits” to which they are entitled
under the VA’s medical program. Powers, 163 F.4th at 1183
(emphasis added).
Under Plaintiffs’ and the panel’s own view of the case,
Plaintiffs’ “meaningful access” claim is that the VA—in
order to avoid discriminating against them in violation of
§ 504—must not only provide them with the medical care
that the program at issue comprises, but also must undertake
a more than $100 million project to build permanent
supportive housing for them. It is unfathomable how the
panel can characterize the requirement to add a completely
different ancillary benefit as not “alter[ing]” the medical care
program, but “simply expand[ing] it.” Powers, 163 F.4th at
5
The majority correctly did not rely on the view that the permanent
supportive housing Plaintiffs seek counts as “domiciliary care”
authorized to be provided under 38 U.S.C. § 1710(b); see also id.
§ 1701(7); 38 C.F.R. §§ 17.30(b), 17.46(b).
POWERS V. MCDONOUGH 17
1195. Under any conceivable standard, the creation of this
sort of new ancillary benefit program is a “fundamental or
substantial modification[]” of the existing program, and is
“far more than the reasonable modifications” required by
§ 504. Choate, 469 U.S. at 300; see also Townsend, 328
F.3d at 518 (stating that a “fundamental alteration” includes
“creat[ing] new programs that provide heretofore
unprovided services to assist disabled persons”). If seeking
to prevent a “reduction from 20 to 14 in the number of
[funded] inpatient hospital days per fiscal year,” Choate, 469
U.S. at 289, amounted to a fundamental alteration of the
underlying program in Choate (and therefore could not be
required as an accommodation), it follows a fortiori that
creating an obligation to provide permanent housing, where
no such obligation existed before, likewise cannot properly
be considered a reasonable accommodation. See Where Do
We Go, 32 F.4th at 862–63 (holding that Caltrans program
to clear homeless encampments and connect homeless
persons with services would be fundamentally altered, and
not merely subject to a “reasonable modification[],” by an
injunction that “effectively asked Caltrans to house Plaintiffs
on its property until Plaintiffs found new housing”).
2
To the extent that the panel’s “reasonable
accommodation” holding implicitly rests on the VA’s
management of other programs for veterans, including
programs involving the provision of housing, that too is legal
error.
As the panel noted, the VA has undertaken the
construction of supportive housing for veterans on the
Campus under other authorities, including the West Los
Angeles Leasing Act of 2016 (“WLALA”), Pub. L. No. 114-
18 POWERS V. MCDONOUGH
226, 130 Stat. 926 (2016), as amended by the West Los
Angeles VA Campus Improvement Act of 2021, Pub. L. No.
117-18, 135 Stat. 288 (2021). See Powers, 163 F.4th at
1174–75. The panel opinion suggests that, in light of this
separate VA housing program, the district court’s judgment
is within the range of “reasonable modifications” that may
be required by § 504 under Choate, 469 U.S. at 300. See
Powers, 163 F.4th at 1195 (holding that the district court’s
housing injunction would not “fundamentally change [the
VA’s] program”). This contention fails for two main
reasons.
First, as explained earlier, the relevant program here for
purposes of determining whether Plaintiffs are being
discriminatorily excluded from “the benefits of . . . any
program or activity conducted by” the VA, 29 U.S.C.
§ 794(a), is the VA’s medical program, not its separate
housing program. See supra at 13–14. Plaintiffs do not
contend that they are being discriminatorily excluded from
any housing program; rather, their contention is that they
need “permanent supportive housing on or near the Campus”
in order to avoid being discriminatorily excluded from
receiving medical care at the Campus. Powers, 163 F.4th at
1183. Indeed, in holding that Plaintiffs’ claims were not
barred by the jurisdiction-stripping provisions of the
Veterans Judicial Review Act, 38 U.S.C. § 511, the panel
placed loadbearing weight on the fact that the relevant
“benefits” were medical benefits, and not housing:
“Plaintiffs do not argue that housing is a benefit, nor have
Plaintiffs requested housing from the [VA] Secretary
through an administrative process. Rather, Plaintiffs
contend that housing is a reasonable accommodation that is
necessary for them to access the benefits the VA has already
conferred to them,” i.e., their medical benefits. Powers, 163
POWERS V. MCDONOUGH 19
F.4th at 1183 (emphasis added). On top of that, the panel
(correctly) held that the “the Leasing Act and its 2021
Amendment” did not create any “judicially enforceable
fiduciary duties on the VA” to build housing. Id. at 1991.
Any housing efforts that are separately undertaken by the
VA are simply not the relevant “program” for purposes of
analyzing Plaintiffs’ claim that, in violation of § 504, they
are being discriminatorily excluded from accessing their
medical benefits. And given that the relevant program is the
medical program, the district court’s injunction obviously
effects—in violation of Choate’s limitations—an
impermissible “fundamental or substantial modification” of
that medical program by adding an ancillary housing benefit
to it. Choate, 469 U.S. at 300.
Second, even assuming arguendo that the two programs
could properly be considered together for purposes of
assessing whether the medical program violates § 504, the
panel’s decision still upholds a substantial alteration of the
housing program that unquestionably counts as a
“fundamental” or “substantial” modification, and that
therefore exceeds what can be required under § 504. The
district court greatly expanded the size of the housing
program and the funding that must be allocated to it, which
further confirms that it effectuates a fundamental alteration.
See Arc of Wash. State Inc. v. Braddock, 427 F.3d 615, 620–
22 (9th Cir. 2005) (holding that requiring an expansion in the
number of persons served by a state program providing
certain services to the disabled would constitute an
impermissible fundamental alteration). 6 Once again, if an
6
The panel’s decision wrongly downplays the amount of the
expenditures ordered by the district court, noting that the “$100 million”
minimum projected price tag was “only 0.02% of the VA’s $407 billion
20 POWERS V. MCDONOUGH
increase from 14 to 20 days of inpatient coverage counted as
a fundamental change in Choate, see 469 U.S. at 301–06,
then an expansion of the housing program on the scale
ordered here certainly does as well.
* * *
The panel decision’s extraordinary holding that a failure
to provide an ancillary benefit (housing) amounts to
discrimination against disabled persons with respect to the
provision of medical care, is squarely contrary to Supreme
Court and Ninth Circuit precedent. On that basis alone,
rehearing en banc should have been granted.
III
The panel opinion’s treatment of Plaintiffs’ so-called
Olmstead claim under § 504 is also deeply flawed and
warranted en banc rehearing.
“In Olmstead, the Supreme Court interpreted the failure
to provide . . . services in a community-based setting as a
form of discrimination on the basis of disability . . . in
violation of [Title II of] the ADA.” Townsend, 328 F.3d at
516–17. (Although Olmstead was an ADA case, the panel
correctly recognized that its analysis would also apply to the
annual budget for 2024.” See Powers, 163 F.4th at 1195. The panel’s
use of this all-budget-percentage benchmark directly contravenes
Supreme Court authority. See Olmstead, 527 U.S. at 603 (plurality) (“If
the expense entailed . . . is properly measured for reasonableness against
the State’s entire mental health budget, it is unlikely that a State, relying
on the fundamental-alteration defense, could ever prevail.”). Moreover,
the $100 million figure covers, not the cost of the entire injunction, but
only the component involving temporary supportive housing, which is a
fraction of the overall project. As the VA noted in its reply brief, that
$100 million dollar figure alone “would consume nearly the entire
annual homelessness budget” of the VA.
POWERS V. MCDONOUGH 21
similarly worded provisions of the Rehabilitation Act.
Powers, 163 F.4th at 1187 n.11.) Under certain
circumstances, failing to provide such community-based
care “contraven[es] the . . . ADA’s ‘integration regulation,’
which require[s] that public entities administer services ‘in
the most integrated setting appropriate to the needs of
qualified individuals with disabilities.’” Townsend, 328
F.3d at 515 (quoting 28 C.F.R. § 35.130(d)); see also 38
C.F.R. § 15.130(d) (imposing a similar requirement on the
VA under the Rehabilitation Act); 28 C.F.R. § 41.51(d)
(same for general Rehabilitation Act regulations).7 In other
words, a failure to “administer services in the most
integrated setting appropriate,” Townsend, 328 F.3d at 515
(simplified), can supply grounds for a plaintiff’s claim that
he or she was denied benefits based on disability in violation
of the ADA, 42 U.S.C. § 12132, and the Rehabilitation Act,
29 U.S.C. § 794. The Olmstead Court referred to this form
of discrimination as both “unjustified isolation” and
“unjustified institutionalization” “of individuals with
disabilities.” Olmstead, 527 U.S. at 597.
As the Supreme Court made clear, an Olmstead claim is
simply another species of “reasonable accommodation”
claim. The liability-creating discrimination arises because,
“[i]n order to receive needed medical services, persons with
mental disabilities must, because of those disabilities,
relinquish participation in community life they could enjoy
given reasonable accommodations.” Olmstead, 527 U.S. at
601 (emphasis added). And, as with a Choate meaningful
access claim, an Olmstead claim cannot be invoked to
7
The panel therefore erred to the extent that it concluded that the relevant
integration regulation here was 28 C.F.R. § 35.130(d). See Powers, 163
F.4th at 1187.
22 POWERS V. MCDONOUGH
require a “fundamental alteration” of the relevant program.
Id. at 603 (four-Justice plurality); id. at 607 (Stevens, J.,
concurring in part and in the judgment) (agreeing with the
plurality on this point).
For the same reasons discussed earlier, Plaintiffs’
housing-construction theory entails a fundamental alteration
of the relevant medical-benefits program, and it therefore
cannot be sustained under Olmstead either. See Townsend,
328 F.3d at 518 (holding that Olmstead does not require
“public entities . . . to create new programs that provide
heretofore unprovided services”).
But there is a further respect in which Plaintiffs’
Olmstead claim fails as a matter of law. The panel’s decision
improperly extends Olmstead to address an entirely different
category of institutionalization risks from those underlying
an Olmstead claim. The gravamen of an Olmstead claim is
that, because the agency only provides the relevant services
in an institutional setting, “[i]n order to receive needed
medical services, persons with mental disabilities must,
because of those disabilities, relinquish participation in
community life they could enjoy given reasonable
accommodation.” Olmstead, 527 U.S. at 601. The risk of
institutionalization underlying an Olmstead claim is thus the
institutionalization associated with receiving the services
offered by the agency. But that is not the “risk of
institutionalization” on which the panel relied. Instead, the
panel grounded the Olmstead claim in the extrinsic risk of
institutionalization associated with the “never-ending cycle
of jail and hospitalization” resulting from their
homelessness. Powers, 163 F.4th at 1187; see also id.
(relying on evidence that homelessness leads to an
“institutional circuit” in which persons end up in “hospitals,
jails and prisons”). Because Plaintiffs’ Olmstead claim does
POWERS V. MCDONOUGH 23
not involve a claim that the VA will institutionalize them in
the course of providing services, but instead relies on
extrinsic risks of institutionalization by others, that claim
lacks the essential ingredient of an Olmstead claim and fails
as a matter of law for that additional reason.
IV
There is a further respect in which both Plaintiffs’
meaningful access claim and their Olmstead claim
contravene settled law. On its face, § 504(a) of the
Rehabilitation Act’s prohibition requires a showing that the
plaintiff was denied a federal benefit “solely by reason of her
or his disability.” 29 U.S.C. § 794(a) (emphasis added).
Plaintiffs’ housing-construction claims plainly fail this
requirement as a matter of law.
We have recognized that “[t]he causal standard for the
Rehabilitation Act is . . . strict[], demanding that [a plaintiff]
show that she was denied services ‘solely by reason of’ her
disability.” Martin v. California Dep’t of Veterans Affs., 560
F.3d 1042, 1049 (9th Cir. 2009); see also, e.g., K.M. ex rel.
Bright v. Tustin Unified Sch. Dist., 725 F.3d 1088, 1099 (9th
Cir. 2013) (“The causal standard for the Rehabilitation Act
is even stricter [than that contained in Title II of the ADA],
requiring a plaintiff to show a denial of services ‘solely by
reason of’ disability.” (simplified)). Here, the multiple
programmatic considerations that inhere in any decision
whether to undertake additional forms of housing
construction, together with the wide variety of life
circumstances that confront the class members, refute any
notion that the VA’s failure to undertake such a program can
be said to be caused “solely by reason of [Plaintiffs’]
disabilit[ies].” 29 U.S.C. § 794(a) (emphasis added).
24 POWERS V. MCDONOUGH
Plaintiffs’ meaningful access and Olmstead claims both fail
as a matter of law for lack of the requisite causal link.
Our decision in Weinreich v. Los Angeles County
Metropolitan Transportation Authority, 114 F.3d 976 (9th
Cir. 1997), confirms this conclusion. In that case, the
plaintiff, Robert Weinreich, sought exemption from a rule of
the Los Angeles County Metropolitan Transportation
Authority (“MTA”) requiring him to continually recertify his
disability in order to participate in a fare-reduction program.
Id. at 978. Weinreich argued that the recertification
requirement violated § 504 “on the ground that he is indigent
and cannot afford to pay a private doctor to recertify his
disability.” Id. We rejected Weinreich’s § 504 claim,
reasoning that he did not show discrimination “solely by
reason of [his] disability.” Id. at 978–79 (citation omitted).
Rather, Weinreich’s exclusion from the fare-reduction
program was based at least in part on “his inability to satisfy
a condition of eligibility because of his financial
circumstances.” Id. at 979. Since Weinreich could not satisfy
§ 504’s “sole[]” causation requirement, the MTA was “not
required under the . . . Rehabilitation Act to make reasonable
modifications to the Program’s eligibility requirements.” Id.
Similar reasoning applies here. Plaintiffs allege that they have
“difficulty accessing healthcare when they don’t have a stable
place to live,” Powers, 163 F.4th at 1186, because
homelessness makes tasks like keeping appointments and
staying in contact with doctors more difficult. As in
Weinreich, Plaintiffs’ inability to access the VA Center does
not flow “solely” from the VA’s failure to build the desired
housing, but from the many individualized problems that each
class member confronts. See also Martin, 560 F.3d at 1048–
49 (upholding verdict rejecting Rehabilitation Act § 504
claim where the denial of services was attributable to multiple
POWERS V. MCDONOUGH 25
factors, including a “lack of resources and space to provide
the level of care that [the plaintiffs’ decedent] required”).
V
For reasons that should be apparent from the foregoing
discussion, the panel also erred in upholding the district
court’s certification of a class under Federal Rule of Civil
Procedure 23(b)(2).
The irreducible minimum of a valid class is “the capacity
of a class-wide proceeding to generate common answers apt
to drive the resolution of the litigation.” Wal-Mart Stores,
Inc. v. Dukes, 564 U.S. 338, 350 (2011) (citation omitted).
This requirement, referred to as commonality, “requires the
plaintiff to demonstrate that the class members ‘have
suffered the same injury.’” Id. at 349–50 (citation omitted).
It is from the class members’ shared injury that “the capacity
of a class-wide proceeding to generate common answers”
arises. Id. at 350 (citation omitted) (emphasis altered).
Correspondingly, “[d]issimilarities within the proposed
class are what have the potential to impede the generation of
common answers.” Id. (citation omitted).
The district court certified a Rule 23(b)(2) class defined
as “[a]ll homeless veterans with Serious Mental Illness
[SMI] or Traumatic Brain Injuries [TBI], who reside in Los
Angeles County.” In upholding the district court’s
conclusion that this class satisfied the commonality
requirement, the panel relied on the assertedly common
question of whether, as a result of the “VA’s failure to
provide supportive housing on or near the Campus,”
“members of Plaintiffs’ putative class are denied meaningful
access to necessary healthcare services and placed at serious
risk of institutionalization.” Powers, 163 F.4th at 1185. But
whether any individual class member needs permanent
26 POWERS V. MCDONOUGH
supportive housing as a reasonable accommodation in
connection with the VA’s provision of medical services
plainly entails an individualized inquiry into the particular
circumstances of each class member. See, e.g., Zukle v.
Regents of Univ. of Cal., 166 F.3d 1041, 1048 (9th Cir. 1999)
(“Whether a particular accommodation is reasonable
depends on the circumstances of the individual case.”
(citation omitted)).
Indeed, a cursory survey of even the named Plaintiffs
reveals sharp dissimilarities among class members, as well
as direct admissions that Plaintiffs have not “suffered the
same injury.” Dukes, 564 U.S. at 350. Most fundamentally,
although the class definition focuses on shared mental-health
conditions, many Plaintiffs have distinct medical conditions
that seem likely to require tailored treatment. By way of a
few brief examples, one named Plaintiff has the disability of
“PTSD, a result of . . . trauma experienced during his Army
service,” as well as “service-connected [] sleep apnea and
tinnitus”; another named Plaintiff’s disabilities include
“multiple sclerosis,” in addition to “Bipolar II, PTSD,
polysubstance use disorder, anxiety, and depression”; and
another unnamed Plaintiff has the disability of “severe
migraine headaches and tinnitus.”
Additionally, multiple named Plaintiffs lived on the
Campus at the time of trial—including one who had elected
to cease his medical treatment with the VA despite his
continued residence there—while others did not. Two
Plaintiffs “leas[ed] a housing unit with assistance from” a
VA-provided voucher, while the others did not. And moving
beyond the named Plaintiffs, the VA has other medical
facilities within the class-defining Los Angeles County, near
which some class members may currently reside.
POWERS V. MCDONOUGH 27
It is evident that not every individual class member will
require, as a reasonable accommodation, permanent housing
next to the Center to the same extent, or in the same fashion,
in order to have “meaningful access” to VA-provided
healthcare. Under the class definition the panel approved,
one class member might have “severe migraine headaches
and tinnitus,” receive a VA-provided housing voucher, and
live near another VA medical facility, while another might
have “multiple sclerosis” and “Bipolar II” and live far from
any VA facility. Although both would equally fall under the
approved class definition, it is obvious that the inquiry into
what is required to provide each class member with
“meaningful access” to healthcare will “depend[] on the
individual circumstances of each case, and require[] a fact-
specific, individualized analysis.” Mark H., 620 F.3d at
1098 (simplified).
* * *
For the foregoing reasons, the panel’s decision
disregards the limitations that the Supreme Court has placed
on disability discrimination claims asserted under § 504 of
the Rehabilitation Act. We should have reheard this
important case en banc. I dissent from our failure to do so.
28 POWERS V. MCDONOUGH
BUMATAY, Circuit Judge, joined by VANDYKE and
TUNG, Circuit Judges, dissenting from denial of rehearing
en banc:
Judges cannot replace their black robes with yellow hard
hats, bulldozing over the clear dictates of the law. Defying
plain text and class action procedure, the district court here
ordered an injunction that requires an unrecognizable, large-
scale transformation of the Department of Veterans Affairs’
(“VA”) provision of medical services to disabled veterans on
its West Los Angeles Campus. See Powers v. McDonough,
163 F.4th 1162, 1180 (9th Cir. 2025). Under the injunction’s
unprecedented terms, the VA must build thousands of
housing units and spend $100 million—nearly all of the
money it had budgeted to tackle homelessness in the greater
Los Angeles area. See id. at 1195. All this to comply with
a district court’s whims as to how to provide “meaningful
access” to the VA’s services for every homeless veteran with
severe mental illness or traumatic brain injuries in the most
populous county in the nation. Id. at 1186–88.
And the VA must build this housing under the district
court’s continuing micromanagement—dictating everything
from where to place portable toilets to where to install a new
pickleball court. See Hr’g on Inj. Relief Tr. at 15:9–16, Dkt.
349, Powers v. McDonough, No. 22-cv-08357 (C.D. Cal.
Oct. 8, 2024) (“[T]here’s no reason that I even have to hook
up to sewer. I can run portable toilets in and showers in and
I prefer to do that on a temporary basis and get people off
the streets. . . . [I]f we can put 40 or 50 or 60 modular 400-
square-foot on that acreage, then there’s no reason that we
can’t bring in temporary showers and temporary toilets and
get these folks out of the rain and then we can worry about
the sewer problem.”); see also Hr’g on Inj. Relief Tr. at
10:9–12, Dkt. 327, Powers, No. 22-cv-08357 (C.D. Cal.
POWERS V. MCDONOUGH 29
Oct. 2, 2024) (“Over the weekend, I’ve had some thoughts
whether [certain tennis courts] might be converted to one
pickleball court [because] maybe the veterans are a lot older
and they can’t play tennis[.]”). Simply, this is preposterous.
Courts are not “roving commissions whose aim is to ‘do the
right thing.’” John Choon Yoo, Who Measures the
Chancellor’s Foot? The Inherent Remedial Authority of the
Federal Courts, 84 Calif. L. Rev. 1121, 1140–41 (1996).
Judge Collins provides a persuasive explanation as to
why Plaintiff-Appellees’ claims should fail on the merits.
But the panel’s errors are even more fundamental. First, in
affirming the district court’s expansive injunction, the panel
took as given that the Rehabilitation Act includes an implied
right of action. See Powers, 163 F.4th at 1186–88
(approving a so-called “Meaningful Access Claim” and
other claims under the Rehabilitation Act without any
reference to statutory text). But no such right exists under
the plain text of the law—instead, we created one out of
whole cloth. See Doe v. Att’y Gen., 941 F.2d 780, 786–87
(9th Cir. 1991), abrogated in part by Lane v. Pena, 518 U.S.
187 (1996). We should have taken this opportunity to
correct that error. Second, the panel reduced the class-action
commonality requirement under Rule 23(a)(2) to a mere
pleading exercise. The panel affirmed commonality by
simply asking whether the VA’s actions “affect[]” the
putative class members, Powers, 163 F.4th at 1185—no
matter the significant differences in medical, economic,
personal, and geographic circumstances among the homeless
individuals making up the class. That violated Article III,
not just Rule 23(a)(2).
To begin, the Rehabilitation Act includes no private right
of action to compel federal agencies to provide disabled
persons “meaningful access” to federal programs and
30 POWERS V. MCDONOUGH
activities, whether express or implied. The Rehabilitation
Act prohibits discrimination on the basis of disability “under
any program or activity receiving Federal financial
assistance or under any program or activity conducted by
any Executive agency.” 29 U.S.C. § 794(a). In turn, the Act
establishes that the “remedies, procedures, and rights set
forth in title VI of the Civil Rights Act of 1964 . . . shall be
available to any person aggrieved by any act or failure to act
by any recipient of Federal assistance or Federal provider of
such assistance.” Id. § 794a(a)(2). So the plain text of the
Rehabilitation Act limits remedies in two important ways.
First, while the Act broadly establishes liability for actions
“by any Executive agency,” id. § 794(a), it narrows remedies
against the federal government only to instances when it acts
as a “Federal provider of [funding] assistance,” id.
§ 794a(a)(2). Second, the Act restricts remedies to those
found in Title VI of the Civil Rights Act of 1964. Id. But
Title VI only creates remedies that extend to federal
agencies’ decisions “to terminat[e] or refus[e] to grant or to
continue financial assistance.” 42 U.S.C. § 2000d-2. Thus,
“[a]lthough the Rehabilitation Act prohibits discrimination
by executive agencies acting as both grant providers and
regulators, it authorizes an express private right of action to
enforce only the grant-making provision.” Moya v. U.S.
Dep’t of Homeland Sec., 975 F.3d 120, 128 (2d Cir. 2020)
(citing 29 U.S.C. § 794a). On top of this, Congress provided
other methods to enforce the Rehabilitation Act’s anti-
discrimination aims—(1) by directing agency heads to
“promulgate such regulations as may be necessary to carry
out the” prohibition, 29 U.S.C. § 794(a), and (2) by allowing
suits against executive agencies to proceed under the
Administrative Procedure Act (“APA”). See Moya, 975
F.3d at 128. That Congress failed to provide a cause of
POWERS V. MCDONOUGH 31
action for alleged discrimination in federal agencies’
operation of federal programs, but did expressly do so for
agencies’ funding decisions, makes clear that courts
shouldn’t imply such a remedy from the statute. See FS
Credit Opportunities Corp. v. Saba Cap. Master Fund, Ltd,
146 S. Ct. 1546, 1553 (2026).
Relying heavily on legislative history and the apparent
purpose of the Rehabilitation Act, however, the Ninth
Circuit has enacted what Congress did not—a private right
of action against allegedly discriminatory federal programs.
Doe, 941 F.2d at 785–87. But the Supreme Court has
repeatedly “rejected the practice of fashioning rights of
action as we see fit.” Cisco Sys., Inc. v. Doe I, 146 S. Ct.
1882, 1890 (2026) (simplified). And whatever the use of
legislative history, it “is not the law.” Epic Sys. Corp. v.
Lewis, 584 U.S. 497, 523 (2018). It can never supersede the
plain meaning of statutory text. See id. No wonder every
other circuit court that has reviewed this issue has rejected a
private right of action against federal programs under the
Rehabilitation Act. See Moya, 975 F.3d at 127–28; Clark v.
Skinner, 937 F.2d 123, 126 (4th Cir. 1991); Cousins v. Sec’y
of the U.S. Dep’t of Transp., 880 F.2d 603, 610–11 (1st Cir.
1989) (en banc) (Breyer, J.); cf. Smith v. Mich. Dep’t of
Corr., 159 F.4th 1067, 1082 (6th Cir. 2025) (rejecting an
implied cause of action for retaliation under the
Rehabilitation Act). We should have overruled our incorrect
precedent and gotten out of the sordid business of judicial
lawmaking. The era of judicially manufactured causes of
action is over.
Second, expanding on a single sentence of an outdated
and little-reasoned circuit precedent, the panel short-
circuited any meaningful analysis of the class-action
commonality requirement. See Powers, 163 F.4th at 1185
32 POWERS V. MCDONOUGH
(relying on Armstrong v. Davis, 275 F.3d 849 (9th Cir.
2001), abrogated in part on other grounds by Johnson v.
California, 543 U.S. 499 (2005)). Armstrong dealt with a
class of disabled prisoners and parolees suing a state parole
authority for disability accommodations during their parole
proceedings. 275 F.3d at 855–57. In a fly-by-night analysis,
the Armstrong panel decided that Federal Rule of Civil
Procedure 23(a)(2)’s commonality requirement was satisfied
because the plaintiff class had made a civil-rights challenge
to a “system-wide practice or policy that affect[ed] all of the
putative class members.” Id. at 868 (simplified). By doing
so, the Armstrong panel made quick work of any “individual
factual differences among the individual litigants” that might
have existed. Id.
Armstrong’s thin reasoning allowed this panel to dodge
any real analysis as to whether this putative class actually
satisfied Rule 23 commonality. Here, the putative class is
defined as “[a]ll homeless veterans with [serious mental
illness] or [traumatic brain injuries], who reside in Los
Angeles County.” And it sought permanent supportive
housing on the VA’s West Los Angeles campus to
accommodate its members’ medical needs. The panel
parroted Armstrong and upheld certification on the grounds
that, “in civil-rights suits,” commonality “is satisfied ‘where
the lawsuit challenges a system-wide practice or policy that
affects all of the putative class members.’” Powers, 163
F.4th at 1185 (quoting Armstrong, 275 F.3d at 868). The
panel then made the bare assertion that the VA’s failure to
build housing on its West L.A. Campus was such a “system-
wide practice or policy.” See id. The panel called it a day
without considering that the members of this expansive class
would require vastly different reasonable accommodations
under the Rehabilitation Act.
POWERS V. MCDONOUGH 33
This can’t be right. It would defy reason to assert that
thousands of disabled homeless veterans in Los Angeles
County each lacked “meaningful access” to the West L.A.
Campus for the same reason—if they lacked access at all.
For each plaintiff, was it discrimination against disability
that prevented them from accessing the VA’s services (as
Plaintiff-Appellees contend)? Or was it something else—
such as the hours-long bus ride to get to the West L.A.
Campus from some parts of Los Angeles? All these
differences are buried under the panel’s approach—just label
a government action “a system-wide practice or policy,”
read “affects” as broadly as possible, and voilà, a class
appears. The result—Rule 23 commonality is satisfied
whenever any diverse set of individuals is “affected” by a
policy, no matter how remotely, tenuously, or abstractly.
This rule flouts basic principles of Article III standing,
violates the Supreme Court’s decision in Wal-Mart Stores,
Inc. v. Dukes, 564 U.S. 338 (2011), and ignores Armstrong’s
distinctive factual context.
Each of these errors is sufficient to require reversal.
Because we should leave the creation of private causes of
action to Congress, and because we must strictly adhere to
the commonality requirement for class actions, I respectfully
dissent from the denial of rehearing en banc.
I.
Private Right of Action
The Rehabilitation Act of 1973, Pub. L. No. 93-112, 87
Stat. 355 (codified as amended at 29 U.S.C. § 701 et seq.),
is a landmark piece of civil rights legislation. One of the
Rehabilitation Act’s many laudable purposes is to “empower
individuals with disabilities to maximize employment,
economic self-sufficiency, independence, and inclusion and
34 POWERS V. MCDONOUGH
integration into society.” 29 U.S.C. § 701(b)(1). The
Rehabilitation Act promises that Americans will not be
denied access to programs on the basis of disability:
No otherwise qualified individual with a
disability in the United States, . . . shall,
solely by reason of her or his disability, be
excluded from the participation in, be denied
the benefits of, or be subjected to
discrimination under any program or activity
receiving Federal financial assistance or
under any program or activity conducted by
any Executive agency or by the United States
Postal Service.
Id. § 794(a). The Rehabilitation Act expressly tasks federal
agencies with promulgating regulations to carry out its anti-
discrimination mandates, id., authorizes the Secretary of
Education to offer technical funds or assistance to public
agencies and nonprofits to remove barriers confronted by
those with disabilities, see id. §§ 705(33), 794b(a), (c), and
provides “aggrieved” employees and other persons with
private causes of action for disability-based discrimination
in employment, see id. 794a(a)(1).
But the Rehabilitation Act does not create a private right
of action that would allow litigants to sue federal agencies
for disability discrimination where those agencies are
running their own operations or providing other services.
That much is clear from an examination of the section of the
Rehabilitation Act that expressly deals with remedies. That
section creates a remedial scheme against entities receiving
federal funds, as well as against federal agencies in their
funding capacities:
POWERS V. MCDONOUGH 35
The remedies, procedures, and rights set forth
in title VI of the Civil Rights Act of 1964 (42
U.S.C. 2000d et seq.) (and in subsection
(e)(3) of section 706 of such Act (42 U.S.C.
2000e–5), applied to claims of discrimination
in compensation) shall be available to any
person aggrieved by any act or failure to act
by any recipient of Federal assistance or
Federal provider of such assistance under
section 794 of this title.
29 U.S.C. § 794a(a)(2). Nothing in the text permits a litigant
to sue the federal government for alleged discrimination
where the government itself provides the services.
That’s decisive. In our system of separated powers,
“private rights of action to enforce federal law must be
created by Congress.” Alexander v. Sandoval, 532 U.S. 275,
286 (2001). After all, only “the people’s elected
representatives” are suited to the “job of resolving how best
to weigh th[e] competing costs and benefits” of private
enforcement actions. Medina v. Planned Parenthood S. Atl.,
606 U.S. 357, 369 (2025). As with other laws, a court’s task
is to interpret these provisions, not to amend them. See
Hernandez v. Mesa, 589 U.S. 93, 100–01 (2020).
The key inquiry is whether the statute itself grants “a
private remedy.” Alexander, 532 U.S. at 286. That’s an
exercise in statutory interpretation: If a statute “does not
displa[y] an intent to create a private remedy, then a cause of
action does not exist and courts may not create one, no
matter how desirable that might be as a policy matter, or how
compatible with the statute.” Ziglar v. Abbasi, 582 U.S. 120,
133 (2017) (simplified). “When Congress creates a private
right of action, it usually does so expressly.” FS Credit
36 POWERS V. MCDONOUGH
Opportunities Corp., 146 S. Ct. at 1552. And given that
courts are no longer permitted to conjure up remedies by
gesturing at statutory purpose, see Alexander, 532 U.S. at
287, any private right of action that is not explicitly granted
must be fairly implied by the statute’s text.
Here, no one argues that the Rehabilitation Act contains
an express right of action against the federal government for
discrimination in running its programs. So Plaintiff-
Appellees’ only hope is to demonstrate that the statutory text
reveals an implied right of action. But that’s a particularly
steep showing given that the Rehabilitation Act was enacted
under the Spending Clause, see Barnes v. Gorman, 536 U.S.
181, 189 n.3 (2002), which means it is “especially unlikely”
to “confer an enforceable right,” Medina, 606 U.S. at 369.
In fact, the Rehabilitation Act’s text shows Congress
intended to confine private remedies to limited
circumstances. Where the Rehabilitation Act does provide
express private rights of action, it does so for only two
specific types of discrimination. The first relates to
employment discrimination. See 29 U.S.C. § 794a(a)(1).
Notably, the provision applies sweepingly to “any
complaint” by “any employee or applicant for employment.”
Id. (emphasis added). The second applies to persons
aggrieved by funding decisions. See id. § 794a(a)(2).
Unlike the employment provision contained in § 794a(a)(1),
the funding provision is circumscribed. As stated above,
under § 794a(a)(2), persons aggrieved “by any recipient of
Federal assistance or Federal provider of such assistance”
can invoke “[t]he remedies, procedures, and rights set forth
in title VI of the Civil Rights Act.”
We know that § 794a(a)(2)’s private right of action is
limited to the federal government’s funding decisions—not
POWERS V. MCDONOUGH 37
its general programs and activities—for two reasons. First,
in contrast to § 794(a), which broadly prohibits
discrimination in “any program or activity conducted by any
Executive agency,” § 794a(a)(2)’s remedies apply narrowly
to only the federal government’s role as “Federal provider of
such assistance.” This difference in language means
something. As the Supreme Court has said in the context of
the waiver of sovereign immunity,
The language of § [794a](a)(2), the remedies
provision, is telling. In that section, Congress
decreed that the remedies available for
violations of Title VI would be similarly
available for violations of § [794](a) “by any
recipient of Federal assistance or Federal
provider of such assistance.” 29 U.S.C.
§ 794a(a)(2). This provision makes no
mention whatsoever of “programs or
activities conducted by any Executive
agency,” the plainly more far-reaching
language Congress employed in § [794](a)
itself. Whatever might be said about the
somewhat curious structure of the liability
and remedy provisions, it cannot be disputed
that a reference to “Federal providers” of
financial assistance in § [794a](a)(2) does
not, without more, establish that Congress
has waived the Federal Government’s
immunity against monetary damages awards
beyond the narrow category of § [794](a)
violations committed by federal funding
38 POWERS V. MCDONOUGH
agencies acting as such—that is, by “Federal
providers.”
Lane, 518 U.S. at 192–93 (simplified). Simply, if Congress
wanted remedies for all § 794(a) violations by the federal
government, it could have easily adopted the sweeping
language of § 794a(a)(1) or the broader description of the
federal government in § 794(a). See id. at 193. Instead, we
have a narrow provision of claims against the federal
government as “Federal provider” of funds. Thus, the Court
interpreted § 794a(a)(2) as applying only to the federal
government’s “funding activities.” Id. at 195.
Second, § 794a(a)(2) only grants the remedies of “title
VI of the Civil Rights Act of 1964 (42 U.S.C. 2000d, et
seq).” 29 U.S.C. § 794a(a)(2). Title VI “preserves whatever
general right to judicial review other laws provide,” Dorsey
v. U.S. Dep’t of Lab., 41 F.3d 1551, 1554 (D.C. Cir. 1994),
but it otherwise only extends judicial review to agency
actions “terminating or refusing to grant or to continue
financial assistance.” 42 U.S.C. § 2000d-2. That’s it. It
creates no private rights of action for other government
activities. Cf. Jersey Heights Neighborhood Ass’n v.
Glendening, 174 F.3d 180, 191 (4th Cir. 1999) (holding that
42 U.S.C. § 2000d-1 does not create a private right of
action); Dorsey, 41 F.3d at 1554 (holding that 42 U.S.C.
§ 2000d does not waive sovereign immunity for money
damages claims). In short, nothing in Title VI creates a
remedy against the federal government for its programs,
activities, and operations unrelated to its grant-making
function.
And the text of the Rehabilitation Act gives another
indication that there’s no such private right of action. Recall
that when Congress creates a detailed remedial scheme, that
POWERS V. MCDONOUGH 39
strongly suggests courts should not recognize remedies that
were not expressly provided for, even where there is
substantive rights-creating language. Alexander, 532 U.S. at
290. That makes sense—while the Supreme Court has
instructed us to assess whether “a statute use[s] ‘rights-
creating language’ aimed at protecting ‘a particular class of
persons,’” FS Credit Opportunities Corp., 146 S. Ct. at 1553
(simplified), such language is “necessary” but not
“sufficient” to infer congressional intent to create a private
judicial remedy, In re Wild, 994 F.3d 1244, 1255 n.11 (11th
Cir. 2021) (en banc) (Newsom, J.) (simplified); accord Allco
Renewable Energy Ltd. v. Mass. Elec. Co., 875 F.3d 64, 71
(1st Cir. 2017).
Such a remedial scheme is present here. For one thing,
the Rehabilitation Act creates a regulatory framework for
federal agencies to comply with its anti-discrimination
mandates. 29 U.S.C. § 794(a); see, e.g., 28 C.F.R.
§ 35.130(d) (requiring public entities to provide their
services “in the most integrated setting appropriate to the
needs of qualified individuals with disabilities”). So
Congress wanted the Executive branch to police itself and
others in enforcing the Act’s anti-discrimination design.
Next, Congress provided a vehicle to sue Executive agencies
through the APA, allowing aggrieved persons to challenge
agency actions contrary to law. See 5 U.S.C. § 706. And
Congress allows the Secretary of Education to offer
technical expertise and funding to States, agencies, and non-
profits to reduce “barriers” faced by disabled persons. 29
U.S.C. § 794b(a), (c). Thus, while the Rehabilitation Act
contains rights-granting language, see id. § 794(a), those
rights are channeled through a separate remedial scheme.
It’s a given in statutory interpretation that “to express or
include one thing implies the exclusion of the other.”
40 POWERS V. MCDONOUGH
Expressio unius est exclusio alterius, Black’s Law
Dictionary (12th ed. 2024). So the fact that Congress
contemplated multiple enforcement mechanisms for the
Rehabilitation Act, including express causes of action for
two specific scenarios (discrimination in employment and
funding), reinforces that Congress did not create a private
cause of action for discrimination when the government
operates federal programs. And the Court has instructed us
that “language establishing an express remedial scheme
elsewhere in the statute may foreclose a private cause of
action to enforce even those statutes that admittedly create
substantive private rights.” FS Credit Opportunities Corp.,
146 S. Ct. at 1553 (simplified). So, “[i]n the absence of
strong indicia of a contrary congressional intent,” we must
“conclude that Congress provided precisely the remedies it
considered appropriate”—no more, no less. Middlesex Cnty.
Sewerage Auth. v. Nat’l Sea Clammers Ass’n, 453 U.S. 1, 15
(1981). The statute is clear. Its remedies do not include a
private right of action for disability discrimination against
federal agencies administering federal programs.
Many circuits have come to a similar conclusion. More
than 35 years ago, then-Judge Stephen Breyer held for the
First Circuit—sitting en banc for good measure—that the
Rehabilitation Act’s text precluded a private cause of action
against the Department of Transportation’s “refusal to
amend, modify, or waive its regulation preventing [the deaf
plaintiff] from driving a truck.” Cousins, 880 F.2d at 605.
Then-Judge Breyer observed that the Rehabilitation Act “is
silent about whether and how a person injured by the
government as regulator is to enforce the Act against the
government.” Id. In concluding that the Rehabilitation Act
contains no implied private right of action for such
circumstances, he reasoned that since judicial review is
POWERS V. MCDONOUGH 41
already generally available against federal agencies under
the APA, there was no need to “strain” to find an implied
cause of action in the Rehabilitation Act. Id. at 610.
A few years later, the Fourth Circuit ruled similarly. As
in Cousins, the Fourth Circuit dealt with the issue in the
context of a suit brought by a disabled truck driver against
the Department of Transportation. Clark, 937 F.2d at 124–
25. And, again like the First Circuit, the Fourth Circuit
rejected finding an implied cause of action under the
Rehabilitation Act. It concluded that plaintiffs cannot sue
for non-employment discrimination claims against the
federal government where the government acts as a
regulator. Id. at 125. It reasoned that the APA provided
sufficient judicial review for claimants who first exhausted
their administrative remedies. Id. at 126.
More recently, the Second Circuit ruled that two
applicants for citizenship could not sue the government
under the Rehabilitation Act for disability exemptions from
their civics and English testing requirements. Moya, 975
F.3d at 124. The Second Circuit started with the principle
that whether an implied cause of action exists is a matter of
statutory text. Id. at 128. It then considered the lack of a
general cause of action against federal agencies in the
Rehabilitation Act, the existence of alternative remedies
(including express causes of action against federal agencies
as to their “grant-making” capacities), and the availability of
judicial review under the APA against federal agencies. Id.
Considering all this, the Second Circuit concluded that the
“Rehabilitation Act does not evince a ‘clear manifestation of
congressional intent’ to create a private right of action
against executive agencies acting in their regulatory
capacity.” Id. (simplified).
42 POWERS V. MCDONOUGH
We are on the wrong side of this lopsided circuit split.
In 1991, we held that a physician had a cause of action under
the Rehabilitation Act against the Justice Department and the
FBI when they refused to send agents to him for
examinations after it was discovered he had AIDS. Doe, 941
F.2d at 782, 794–95. We recognized that no statutory text
expressly established a cause of action. See id. at 787. And
elsewhere in Doe, we even acknowledged that sovereign
immunity erects a higher bar for implying a private right of
action. Id. at 789. But we did not know when to give up—
we charged forward, wielding floor statements by individual
legislators to assert that “Congress knew” some courts had
interpreted the Act to provide an implied private cause of
action. Id. at 786–87. We also relied on a single question
asked by one Senator to similarly decide that Congress had
intended for the Act to offer a broad remedy for all instances
of disability discrimination. Id. at 787 (“In the congressional
debates on section [794a], Senator Bayh asked and Senator
Cranston confirmed that ‘section [794a] merely extends to
the handicapped the same remedies, procedures and rights
already extended’ through Titles VI and VII of the Civil
Rights Act of 1964 and Title IX of the Education
Amendments of 1972.” (simplified)). We then rode this
swell of assumptions and legislative history to conclude that
“the most reasonable interpretation” of the Rehabilitation
Act was to make the federal government just as liable as
private individuals. Id. at 791; see also id. at 791–93
(extensively examining House and Senate floor debates
among individual legislators). But “at the very best,” this
legislative history was “ambiguous.” Dorsey, 41 F.3d at
1555 (criticizing the Ninth Circuit’s analysis). We even
admitted that “the debates do not state outright that
[§ 794(a)] subjects federal agencies to private actions for
POWERS V. MCDONOUGH 43
money damages,” Doe, 941 F.2d at 792, despite reasoning
that the debates “unequivocally express Congress’s intent to
do precisely that,” id. at 793.
Doe’s reading of the Rehabilitation Act flies in the face
of the Supreme Court’s repeated admonitions not to imply
causes of action where the text does not provide for them,
Cisco Sys., 146 S. Ct. at 1890, and not to rely on individual
legislators’ statements to change the plain meaning of
statutes, NLRB v. SW Gen., Inc., 580 U.S. 288, 307 (2017)
(“[F]loor statements by individual legislators rank among
the least illuminating forms of legislative history.”
(simplified)). So while the Supreme Court has sometimes
heard cases considering implied Rehabilitation Act suits
against private or state entities, see, e.g., Cummings v.
Premier Rehab Keller, P.L.L.C., 596 U.S. 212, 217–18
(2022), we should not extend precedent beyond what the
Court has expressly recognized. Given that “judicially
created causes of action offend the separation of powers in
almost every circumstance,” Cisco Sys., 146 S. Ct. at 1891,
the better course is always to follow the plain text of the law.
See Cummings, 596 U.S. at 230–31 (Kavanaugh, J.,
concurring) (“Congress, not this Court, creates new causes
of action. And with respect to existing implied causes of
action, Congress, not this Court, should extend those implied
causes of action and expand available remedies.”
(simplified)). To do otherwise would aggrandize courts at
the expense of the legislature. It would also undermine
Congress’s carefully crafted remedial schemes, which
balance accountability and sovereignty. The Rehabilitation
Act’s text simply does not permit a private cause of action
against federal agencies operating federal programs. Our
precedent to the contrary should be overruled.
44 POWERS V. MCDONOUGH
Plaintiff-Appellees ask us to ignore this wall of
authority, claiming that Cousins, Clark, and Moya involved
the government acting as a “regulator” rather than as a
provider of services, as here. Similarly, Doe tried to brush
off Cousins by claiming that a distinction exists under the
Rehabilitation Act when the government acts in its
“propriety capacity” rather than in its “administrative or
regulatory capacities.” See 941 F.2d at 793. But this
distinction was created out of thin air. Nothing in the text of
the Rehabilitation Act supports this dividing line: just as
§ 794a(a)(2) says nothing about agencies acting in their
“regulatory” capacities, it says nothing about agencies acting
in their “proprietary” capacities. And in pressing its
distinction, Doe doubled down on the idea that the
Rehabilitation Act was meant to provide for damages claims
against the federal government. See 941 F.2d at 793. That
conclusion—again rooted in freewheeling legislative
history, see id. at 794—was rejected by the Supreme Court
only a few years later. See Lane, 518 U.S. at 197. And for
what it’s worth, our court largely abandoned the
proprietary/regulatory distinction almost immediately. See
J.L. v. Soc. Sec. Admin., 971 F.2d 260, 269 (9th Cir. 1992),
overruled in part by Lane, 518 U.S. 187 (“Doe’s conclusion
that Congress intended to permit private suits against the
government did not turn on the proprietary/regulatory
distinction.”).
Finally, Plaintiff-Appellees argue that the federal
government waived this argument by not pressing it before
the panel. But the federal government sufficiently preserved
this issue for review. While it primarily focused on whether
the Veterans Judicial Review Act of 1988, Pub. L. No. 100-
687, 102 Stat. 4113 (codified as amended at 38 U.S.C.
§ 7251 et seq.), stripped us of subject-matter jurisdiction, the
POWERS V. MCDONOUGH 45
federal government made clear in its opening brief an intent
to challenge in “an appropriate case” this court’s precedent
holding that the Rehabilitation Act provides an implied
private right of action against federal programs. See
Opening Br. 30 n.2 (citing Doe, 941 F.2d at 787). And in its
petition for rehearing en banc, the federal government
confirmed that this is now “an appropriate case.” That’s
because we could not have reconsidered Doe at the panel
stage, so the government’s request to do so would have been
“futile.” United States v. Hernandez-Estrada, 749 F.3d
1154, 1160 (9th Cir. 2014) (en banc). In any event, “we have
the authority and discretion to decide questions first raised
in a petition for rehearing en banc” and “we have done so at
the government’s request.” Id. at 1159–60 (simplified).
Both parties addressed this pure question of law in their en
banc petition briefing, so nothing bars us from considering
the question as an en banc court. See Lee v. Fisher, 70 F.4th
1129, 1154 (9th Cir. 2023) (en banc). Indeed, when the issue
is of great “importance” (as this one no doubt is), we should
feel obliged to do so. Id.
II.
Class-Action Commonality
Class actions are “an exception to the usual rule that
litigation is conducted by and on behalf of the individual
named parties only.” Wal-Mart, 564 U.S. at 348
(simplified). Though class actions can create efficiencies for
courts and parties, they also raise distinct dangers. See, e.g.,
Kincannon v. United Airlines, Inc., 168 F.4th 713, 735 (5th
Cir. 2026) (Willett, J., concurring in part and concurring in
the judgment). Opportunistic litigants hungry for damages
may assert themselves as leaders of classes that didn’t ask to
be led, creating undue pressures on defendants to settle for
46 POWERS V. MCDONOUGH
fear of a ruinous judgment. See, e.g., id.; see also Olean
Wholesale Grocery Coop., Inc. v. Bumble Bee Foods LLC,
31 F.4th 651, 685 (9th Cir. 2022) (en banc) (Lee, J.,
dissenting) (“If a court certifies a class, the potential liability
at trial becomes enormous, maybe even catastrophic, forcing
companies to settle even if they have meritorious
defenses.”). The prospect of injunctive relief against the
government on behalf of large classes can convert courts into
quasi-legislators, effectively allowing judges to make policy
choices on behalf of thousands—or even millions—of
people. See Trump v. CASA, Inc., 606 U.S. 831, 867–68
(2025) (Alito, J., concurring). To that end, we must treat
class certification with care.
So we must meticulously enforce the prerequisites for
maintaining class actions set forth in Federal Rule of Civil
Procedure 23(a). One of those prerequisites is commonality:
class plaintiffs must show that “there are questions of law or
fact common to the class.” Fed. R. Civ. P. 23(a)(2). This
means that there must be a “common contention” whereby
the “determination of its truth or falsity will resolve an
issue . . . in one stroke.” Wal-Mart, 564 U.S. at 350
(simplified). So, to satisfy commonality, the issue isn’t just
whether class plaintiffs raise “common questions,” but
whether they have shown “the capacity of a class-wide
proceeding to generate common answers apt to drive the
resolution of the litigation.” Id. (simplified).
In this case, the certified class was defined as all
“homeless veterans” with serious mental illness or traumatic
brain injuries residing in Los Angeles County. Powers, 163
F.4th at 1179. The putative class asserted that the VA’s
failure to provide housing near its West L.A. Campus
resulted in systematic discrimination against the putative
class members in violation of the Rehabilitation Act. Id. at
POWERS V. MCDONOUGH 47
1185; see also id. at 1179–80. The VA argued that the class
definition failed commonality under Rule 23(a)(2) because
the Rehabilitation Act claims necessarily turned on each
veteran’s individual circumstances and thus the class action
couldn’t generate common answers to resolve the litigation.
Id. at 1184–85. Relying on Armstrong, the panel approved
of the class because “Plaintiffs challenge the VA’s failure to
provide permanent housing on or near the Campus, a system-
wide practice that Plaintiff-Appellees claim is
discriminatory and that affects all of the putative class
members.” Id. (simplified). According to the panel, that was
enough. Id. Any differences between the individual
veterans, the panel said, “merely relate to the degree to
which each class member was denied meaningful access to
healthcare and their risk of institutionalization.” Id. at 1185
n.10.
The panel’s ruling missed the mark in three different
ways. First, it allowed the district court to exceed its own
jurisdiction by certifying a class action in which the class
members did not share common Article III injuries. Second,
it conflicted with the Supreme Court’s clear instructions in
Wal-Mart about how to interpret Rule 23(a)(2). And third,
it expanded Armstrong’s broad language invoking “system-
wide” effects so that the commonality requirement was
stretched beyond its limits. I discuss each in turn.
A.
Article III Standing
The greatest flaw with the panel’s expansive view of
class actions is that it created commonality between injured
and uninjured class members. Two of the most important
checks on our powers as federal judges are that we can only
decide “concrete disputes between an injured plaintiff and a
48 POWERS V. MCDONOUGH
defendant according to the law,” Washington v. Trump, 145
F.4th 1013, 1041 (9th Cir. 2025) (Bumatay, J., concurring in
part and dissenting in part), and “that we may only grant
party-specific relief,” id. at 1040. These checks “prevent the
judicial process from being used to usurp the powers of the
political branches.” Clapper v. Amnesty Int’l USA, 568 U.S.
398, 408 (2013) (simplified). At the end of the day, “[t]he
Founders left non-particularized challenges to disfavored
policy to the ballot box—not the courts.” Washington v.
Trump, 145 F.4th at 1041 (Bumatay, J., concurring in part
and dissenting in part). So “we must resist the temptation to
expand our authority by . . . allowing . . . jurisdictional end-
runs.” Id.
“The federal court system is reserved only for those that
have suffered an injury” cognizable under Article III. Olean
Wholesale Grocery Coop., Inc. v. Bumble Bee Foods LLC,
993 F.3d 774, 791 n.7 (9th Cir. 2021), vacated on reh’g en
banc, 31 F.4th 651 (9th Cir. 2022); see FDA v. All. for
Hippocratic Med., 602 U.S. 367, 380 (2024). A class action
is no different. See Olean Wholesale Grocery Coop., 993
F.3d at 791 n.7; see also Tyson Foods, Inc. v. Bouaphakeo,
577 U.S. 442, 446 (2016) (Roberts, C.J., concurring)
(“Article III does not give federal courts the power to order
relief to any uninjured plaintiff, class action or not.”);
Speerly v. Gen. Motors, LLC, 143 F.4th 306, 352 (6th Cir.
2025) (en banc) (Nalbandian, J., concurring) (determining,
after surveying historical practice, that a district judge who
“certifies a class containing individuals who lack Article III
standing” risks committing an ultra vires act). That’s why
“[t]he presence of uninjured parties in a certified
class . . . raises serious standing implications under Article
III.” Olean Wholesale Grocery Coop., 993 F.3d at 791 n.7.
After all, “if there are members of a class that aren’t even
POWERS V. MCDONOUGH 49
injured, they can’t share the same injury with the other class
members.” Lab’y Corp. of Am. Holdings v. Davis, 605 U.S.
327, 332 (2025) (Kavanaugh, J., dissenting) (simplified). In
sum, Article III places limits on commonality under Rule 23
and prohibits us from approving an expansive class like the
one here.
The class definition of all Los Angeles County disabled
homeless veterans exceeded what Article III permits. The
class necessarily swept in veterans who would refuse
permanent supportive housing if offered or who don’t seek
medical treatment at the West L.A. Campus at all. Also
included were veterans who don’t require any
accommodations because they already reside at or near the
West L.A. Campus.
Take plaintiff Lavon Johnson, a veteran with PTSD (and
other conditions). Mr. Johnson already lives on the grounds
of the West L.A. Campus. He receives regular medical care
at his unit. When asked if living on Campus made it easier
to access healthcare services, he responded that his housing
“put [him] in a place where they can find [him].” With this
testimony, it’s hard to see how he’s an injured member of
the putative class.
Or take plaintiff Laurieann Wright, who illustrates
another standing problem. She suffers from PTSD and lives
in VA-affiliated housing in Lancaster—almost 65 miles
away from the West L.A. Campus. Ms. Wright apparently
could travel to the Campus, but she indicated that sitting in
traffic or on a train for hours is painful because of her
physical disabilities. Thus, it’s not clear that the VA’s
failure to provide housing has solely injured Ms. Wright—
rather, her physical disabilities play a role. So even if she is
injured by the VA’s actions, her claims require a different
50 POWERS V. MCDONOUGH
analysis than those of a veteran without physical disabilities
or a long commute.
Contrary to the panel’s assertion, these individual
circumstances aren’t mere differences in “degree” of injury.
Powers, 163 F.4th at 1185 n.10. Some putative class
members suffered no injury at all. Others may have suffered
injuries but can’t trace them to the VA’s alleged
discriminatory practices. Still others may have injuries not
redressed by the district court’s injunction. The panel
ignored this jurisdictional problem by taking a sweeping
view of what “affects” class members. Id. at 1185. This
conflated effects with injuries. Even if a challenged policy
“affects” an individual, it doesn’t necessarily “injure” that
individual for Article III purposes.
B.
Impact of Wal-Mart
Even more, the panel’s reliance on Armstrong conflicted
with the Supreme Court’s teachings in Wal-Mart. The
government contested the commonality of Plaintiff-
Appellees’ two sets of class claims—the “meaningful access
claims” and the so-called “Olmstead claims,” named for
Olmstead v. L.C. ex rel. Zimring, 527 U.S. 581 (1999). The
panel botched the commonality inquiry for both.
In assessing commonality, Wal-Mart instructs that a
putative class must establish that class members “have
suffered the same injury.” 564 U.S. at 350 (simplified).
Once again, this requires a showing that the class action will
“generate [a] common answer[]” to “an issue that is central
to the validity of each one of the [class members’] claims in
one stroke.” Id. (simplified). That goes well beyond a
showing “merely that [class members] have all suffered a
POWERS V. MCDONOUGH 51
violation of the same provision of law.” Id. And the
showing is necessarily fact-intensive, see id. at 353, given
that “[d]issimilarities within the proposed class . . . have the
potential to impede the generation of common answers,” id.
at 350 (simplified). Confirming commonality is a rigorous
undertaking. See id. at 350–51.
While the panel paid lip service to Wal-Mart, its analysis
was sorely lacking. That analysis hinged on Plaintiff-
Appellees’ allegation that “the VA’s failure to provide
supportive housing on or near [the West L.A. Campus]
results in systemic discrimination against them. As a result,
[class members] are denied meaningful access to necessary
healthcare services and placed at serious risk of
institutionalization.” Powers, 163 F.4th at 1185. The panel
further stated that “Plaintiffs’ claim that the VA’s ‘past and
current practice’ of not providing permanent housing to the
class of veterans represents discrimination based on
disability” satisfies the commonality requirement. Id. at
1185.
Wal-Mart was invoked, but it was not followed. After
all, Wal-Mart was clear: “Rule 23 does not set forth a mere
pleading standard. A party seeking class certification must
affirmatively demonstrate . . . that there are in fact . . .
common questions of law or fact.” 564 U.S. at 350. That
certainly wasn’t done here. Indeed, it’s not even clear
whether this conclusory analysis would survive a basic run-
in with Twombly or Iqbal. That both the meaningful access
and Olmstead claims required fact-intensive showings only
magnifies the panel’s errors.
Start with meaningful access claims. The Supreme
Court has established that “an otherwise qualified
handicapped individual must be provided with meaningful
52 POWERS V. MCDONOUGH
access” to a benefit, and that this may require a reasonable
accommodation. Alexander v. Choate, 469 U.S. 287, 301
(1985). In turn, the benefit itself “cannot be defined in a way
that effectively denies otherwise qualified handicapped
individuals the meaningful access to which they are
entitled.” Id. The panel accepted Plaintiff-Appellees’
position “that housing is a reasonable accommodation that is
necessary for them to access the benefits the VA has already
conferred to them.” Id. at 1183. But the very core of this
assertion depends on the fact-specific, individualized
question of whether this housing is indeed a reasonable
accommodation. See Mark H. v. Hamamoto, 620 F.3d 1090,
1097–98 (9th Cir. 2010). This is nowhere in the opinion.
The same goes for the “Olmstead claims.” Olmstead
provided that:
States are required to provide community-
based treatment for persons with mental
disabilities when the State’s treatment
professionals determine that such placement
is appropriate, the affected persons do not
oppose such treatment, and the placement can
be reasonably accommodated, taking into
account the resources available to the State
and the needs of others with mental
disabilities.
527 U.S. at 607. Transposing that holding over to the present
case, it’s clear that these claims were similarly littered with
factual landmines. That’s the case even if Plaintiff-
Appellees “need only show that the challenged [agency]
action creates a serious risk of institutionalization.” Powers,
163 F.4th at 1187 (quoting M.R. v. Dreyfus, 697 F.3d 706,
POWERS V. MCDONOUGH 53
734 (9th Cir. 2012)); cf. M.R., 697 F.3d at 726 (reversing a
district court that “did not sufficiently consider
individualized evidence” when assessing whether plaintiffs
bringing claims under the Rehabilitation Act would suffer
irreparable injury). But the panel rebuffed the need for
convincing evidence showing commonality. Powers, 163
F.4th at 1185.
Instead, the panel applied Armstrong, which suggests
that commonality is per se established for civil rights actions
when “a system-wide practice or policy . . . affects all of the
putative class members.” 275 F.3d at 868. At least on its
face, that’s more expansive than Wal-Mart’s rule, which
would dictate that any such alleged harm will satisfy
commonality only where its truth or falsity will “resolve an
issue that is central to the validity of each one of the claims
in one stroke.” 564 U.S. at 350. And compare the panel’s
evidence-free analysis with the fact-intensive deep dive that
Wal-Mart conducted in assessing commonality. See id. at
351–58.
In sum, we were told that the lack of permanent housing
on the West L.A. Campus “affects all of the putative class
members.” Powers, 163 F.4th at 1185. But how? What was
the “glue” whereby all class members were affected in a way
that is relevant to the Rehabilitation Act claims? Cf. Wal-
Mart Stores, 564 U.S. at 352. We got none of those answers.
Simply, the VA’s decision not to build permanent housing
on its West L.A. Campus doesn’t injure each homeless
veteran in Los Angeles County in a common way—nor does
it “affect” each homeless veteran in a similar way to answer
whether the VA has discriminated on the basis of disability.
It is not enough for class members to be “affected” by a
policy in the broadest, most expansive sense of that term.
Rather, to comport with Wal-Mart, the challenged “practice
54 POWERS V. MCDONOUGH
or policy” must directly act on all of the class members and
cause common injuries. The panel got it wrong.
C.
Armstrong’s Factual Context
The tensions between Wal-Mart and Armstrong are
obvious. But even if we were to accept Armstrong, the panel
ignored all its context in deeming the VA’s decision not to
build permanent housing on its Campus a “system-wide
practice.” Powers, 163 F.4th at 1185. As we have long said,
the “language of the court must be read in the light of the
facts before it.” Julian Petrol. Corp. v. Courtney Petrol. Co.,
22 F.2d 360, 362 (9th Cir. 1927); accord Marshall v. Andrew
F. Mahony Co., 56 F.2d 74, 78 (9th Cir. 1932). By plucking
Armstrong’s “system-wide practice or policy” language
from its context, the panel created an end run around the
commonality requirement.
In Armstrong, disabled state prisoners and parolees sued
the California Board of Prison Terms and state officials,
alleging that the state parole authority failed to make certain
accommodations for them at parole proceedings in violation
of federal law. Armstrong, 275 F.3d at 854. They alleged
that the Board’s parole notification, hearing, and appeals
process required written forms that they could not
understand without accommodations, which in turn caused
those prisoners and parolees to forfeit their rights or be
unable to adequately represent themselves at their parole
hearings. Id. at 856–58. The class consisted of prisoners
and parolees with six categories of disability: mobility
impairments, hearing disabilities, visual disabilities,
learning disabilities, mental retardation, and renal
impairments. Id. at 854. Though the Board sought to split
the state prisoners into different classes, the Armstrong panel
POWERS V. MCDONOUGH 55
disagreed despite some “individual factual differences”
among the groups. Id. at 868. The panel reasoned that “the
differences that exist here do not justify requiring groups of
persons with different disabilities, all of whom suffer similar
harm from the Board’s failure to accommodate their
disabilities, to prosecute separate actions.” Id.
The “system-wide practice or policy” in Armstrong was
very different than what we have in this case. There, the
prisoners sued the Board for the Board’s failure to
accommodate their ability to protect themselves in
proceedings run by the Board. Thus, the Board had
complete and direct control over the prisoners and the parole
“system.” And the “system” injured the prisoners in the
same relevant way—the lack of accommodations precluded
adequate representation in parole hearings, even if the
precise accommodation would differ among the different
categories of disability. See Armstrong, 275 F.3d at 868–69.
So the defendants directly controlled the “system” by which
the Armstrong plaintiffs defined their class, and the
challenged policy or practice acted directly on each class
members in the same way. In other words, because the
prisoners all suffered from the same harm, the same
remedy—making parole notices more accessible—would
redress their injuries.
In contrast, the panel here didn’t explain how the VA has
“system-wide” control over the class of homeless veterans
or how its limited housing policy on the West L.A. Campus
has constituted a “system-wide” policy or practice. The VA
certainly doesn’t control Los Angeles’s county-wide
housing or population of disabled homeless veterans. Far
from it. Indeed, it’s likely that significant numbers of class
members were outside the reach of the VA’s housing
policy—it doesn’t and can’t house the thousands of
56 POWERS V. MCDONOUGH
homeless veterans in the county. In fact, the panel didn’t
even say how many of the class members actually sought VA
care at the West L.A. Campus or how many of those would
have directly benefitted from permanent housing on the
West L.A. Campus. Indeed, it’s anyone’s guess as to how
the VA’s “systematic” housing policy has “affected” the
putative class—if at all. The panel never told us beyond
superficialities. See Powers, 163 F.4th at 1185. Under the
panel’s approach, any challenged action could be labeled
“systematic” and satisfy Armstrong’s low bar.
Even if Armstrong survives Wal-Mart (which I doubt), it
doesn’t apply here. Plaintiff-Appellees defined their class as
all homeless veterans with serious mental illness or
traumatic brain injuries across all of Los Angeles County,
(rightfully) decried the unique harms that all such homeless
veterans face, and sought permanent supportive housing “on
or near” the VA campus to rectify those harms. The problem
is that the VA does not control every aspect of veteran
homelessness across the County, and its decisions thus won’t
“affect[]” all class members in a similar way. See
Armstrong, 275 F.3d at 868. So any remedies ordered won’t
redress the entire class as they did in Armstrong. Rather, the
order to build permanent housing only helps those homeless
veterans who would actually live on its grounds or otherwise
directly benefit from the housing. Simply, the class was too
broadly defined for Armstrong to automatically confer
commonality.
III.
Conclusion
Many may disagree with the lack of housing for veterans
on the VA’s West L.A. Campus; just as many may also
disagree with the decisions that have led to this situation.
POWERS V. MCDONOUGH 57
But judges cannot solve every problem. We have no
business ordering the VA to build thousands of housing units
on a whim. This is not the Ninth Circuit Court of
Appropriations. And we cannot judicially create a cause of
action when Congress refuses to do so. Nor can we conjure
a class when no commonality exists. Precedent saying
otherwise should be overruled.
I respectfully dissent.