FOR PUBLICATION
  UNITED STATES COURT OF APPEALS
       FOR THE NINTH CIRCUIT

KATHLEEN M. WINN, an Arizona           
taxpayer; DIANE WOLFTHAL, an
Arizona taxpayer; MAURICE
WOLFTHAL, an Arizona taxpayer
LYNN HOFFMAN, an Arizona
taxpayer,                                    No. 05-15754
              Plaintiffs-Appellants,             D.C. No.
                 v.
                                          CV-00-00287-EHC
                                           District of Arizona,
ARIZONA CHRISTIAN SCHOOL TUITION
ORGANIZATION; ARIZONA SCHOOL                     Phoenix
CHOICE TRUST; LUIS MOSCOSO;                      ORDER
GALE GARRIOTT, in his official
capacity as Director of the
Arizona Department of Revenue;
GLENN DENNARD,
             Defendants-Appellees.
                                       
                   Filed October 21, 2009

    Before: Dorothy W. Nelson, Stephen Reinhardt and
            Raymond C. Fisher, Circuit Judges.

                          Order;
             Concurrence by Judge Pregerson;
            Concurrence by Judge D.W. Nelson;
              Dissent by Judge O’Scannlain




                            14703
14704           WINN v. ARIZONA CHRISTIAN SCHOOL
                              ORDER

  Judges Reinhardt and Fisher voted to reject the petitions for
rehearing en banc and Judge Nelson so recommended.

   The full court was advised of the petitions for rehearing en
banc. A judge requested a vote on whether to rehear the mat-
ter en banc, and the matter failed to receive a majority of the
votes of the nonrecused active judges in favor of en banc con-
sideration. Fed. R. App. P. 35.

  The three petitions for rehearing en banc, filed May 14,
2009, are denied.



PREGERSON, Circuit Judge, concurring:

  For the reasons stated by the panel in its concurrence, I also
concur in the denial of rehearing en banc.



D.W. NELSON, REINHARDT and FISHER, Circuit Judges,
concurring in the denial of rehearing en banc:

  A majority of the active judges of our court declined to
vote for rehearing of this case en banc. We concur in the
court’s decision not to go en banc.

   The State of Arizona finances private “school tuition orga-
nizations” (STOs) by giving dollar-for-dollar tax credits to
individuals who contribute to them. On its face, the statute
creating this subsidy requires STOs to provide scholarships
for students “to attend any qualified school of their parents’
choice.” Ariz. Rev. Stat. Ann. § 43-1089(G)(3) (2005)
(emphasis added).1 As the Arizona Department of Revenue
  1
   All references to “Section 1089” refer to the program as set forth in
Arizona Revised Statutes Annotated § 43-1089 (2005), the version of the
                 WINN v. ARIZONA CHRISTIAN SCHOOL                    14705
applies the statute, however, the state reimburses contribu-
tions to STOs that restrict their scholarships to use at religious
schools. Consequently, 85 percent or more of the state-
financed scholarship money is available only to students
whose parents are willing to send them to sectarian institutions.2

   If these facts are proved true, the Arizona Department of
Revenue’s execution of the scholarship program (Section
1089) violates the Establishment Clause. In Zelman v.
Simmons-Harris, 

(2002), the Court upheld a
school voucher program that “provide[d] assistance directly”
to parents without regard for religion, and public funds
reached religious institutions only as the result of parents’
choices about their children’s education. 

Parents
received aid whether or not they were willing to enroll their
children in sectarian schools, so the program did not exert
pressure on parents to choose religious schools. Under the
Arizona program, by contrast, taxpayers, rather than parents,
direct funds to religious organizations. Access to assistance is
restricted on the basis of religion, creating financial incentives
that may skew parents’ choices toward religious schools. See

statute in place when plaintiffs’ complaint was filed. Any differences
between this and the current version of Section 1089 are not significant for
purposes of the analysis.
   2
     The dissent sees no constitutional distinction between a tax deduction
and a tax credit. See Dissent 14721 n.3. We disagree. A tax-credit eligible
contribution to an STO costs the taxpayer nothing. See Winn v. Killian,

, 1015 n.5 (9th Cir. 2002), aff’d sub. nom Hibbs v. Winn,

(2004) (“From a purely financial perspective, . . . a taxpayer
is unaffected by his decision as to whether or not to make an STO contri-
bution. The funds that he may contribute will be unavailable to him in any
event: they will be used either to make the contribution or to pay the taxes
he owes.”). Tax deductible contributions, by contrast, impose a cost on the
taxpayer. See 

taxpayer is entitled to a tax deduction, the
taxpayer must in most if not all instances still pay a majority of the tax
involved[.]”). Whereas a tax deduction would lower the cost of contribu-
tions to STOs, a dollar-for-dollar tax credit reduces that cost to zero, in
effect allowing individual taxpayers to directly allocate public funds.
14706             WINN v. ARIZONA CHRISTIAN SCHOOL

The differences between the Ohio and Arizona pro-
grams are constitutionally meaningful.

   The dissent from rehearing en banc demonstrates that oth-
ers may prefer a more expansive reading of Zelman. Careful
review of the two cases, however, shows why the dissent’s
argument that “Winn cannot be squared with the Supreme
Court’s mandate in Zelman” is not persuasive. Dissent 14719-
20.

                            I.   Background

   On its face, Section 1089 appears to provide for parental
choice. The statute says that for an organization to qualify as
an STO eligible to receive state-reimbursed contributions the
organization must provide scholarships “to children to allow
them to attend any qualified school of their parents’ choice.”
Ariz. Rev. Stat. Ann § 43-1089(G)(3) (emphasis added).3 If
this were how Arizona applied the statute, the Arizona pro-
gram would be similar to the majority of the tax credit pro-
grams currently in operation. Like Section 1089, four of these
programs contain provisions directing scholarship organiza-
tions to provide scholarships for students to attend any quali-
fied school of their parents’ choice.4 These “parental choice”
clauses may explain the apparent absence of any Establish-
ment Clause challenges to those programs.
  3
     A “qualified school” is defined by statute as “a nongovernmental pri-
mary school or secondary school or a preschool for handicapped students
that is located in this state, that does not discriminate on the basis of race,
color, handicap, familial status or national origin and that satisfies the
requirements prescribed by law for private schools in this state.” 

1089(G)(2).
   4
     See Fla. Stat. § 220.187(6)(h); Ga. Code Ann. § 20-2A-1(3)(A); Ind.
Code § 20-51-3-1(b); Iowa Code § 422.11S(5)(c)(1); R.I. Gen. Laws § 44-
62-2(a). One of the currently operating programs contains no such parental
choice provision. See 24 Pa. Stat. Ann. § 20-2005-B.
              WINN v. ARIZONA CHRISTIAN SCHOOL             14707
   This is not, however, how the Arizona Department of Rev-
enue applies the statute. According to plaintiffs’ complaint,
Arizona gives tax credits reimbursing individuals who con-
tribute to STOs that expressly restrict their scholarships to use
at religious schools. The largest STO, the Catholic Tuition
Organization of the Diocese of Phoenix, restricts its scholar-
ships to use at Catholic schools in the Phoenix Diocese; the
second largest, the Arizona Christian School Tuition Organi-
zation provides scholarships only to students attending evan-
gelical Christian schools; and the third largest, Brophy
Community Foundation, restricts its scholarships to use at two
specific Catholic schools. See 

. As a
result of how the Arizona Department of Revenue applies
Section 1089, plaintiffs allege, these three religious STOs
controlled 85 percent of the total STO donations in 1998, the
year before the complaint was filed. See 

Plaintiffs
argue that it is this application of Section 1089 that violates
the Establishment Clause.

                          II.   Effect

   We turn first to the issue of “whether Section 1089 ‘has the
forbidden “effect” of advancing or inhibiting religion.’ ”

(quoting 

).

                    A.    Parental Choice

   The Winn panel held that the Arizona Department of Reve-
nue’s application of Section 1089 may not provide parents
with “true private choice” within the meaning of Zelman. See

With respect to this conclusion, the dissent
accuses the panel of rejecting the majority’s holding in Zel-
man in favor of Judge Souter’s dissent. Not so.

                                1.

   The dissent fails to address the crucial difference between
the Ohio voucher program upheld in Zelman and the Arizona
14708         WINN v. ARIZONA CHRISTIAN SCHOOL
Department of Revenue’s application of Section 1089: with
respect to religion, the Ohio program gave parents equal
access to tuition benefits. See 

. Under
the Ohio program, the state provided tuition aid on the basis
of financial need, without regard to religion, and eligible par-
ents were free to apply the aid toward any private school, reli-
gious or secular, or toward a public school outside the district
willing to participate in the program (though none was). If a
parent decided to send her child to a private school, the state
wrote a check made payable to the parent, which the parent
could then endorse over to her chosen school. See 

Crucially, a parent’s choice to send her child to a religious
school would neither help nor harm her chance of receiving
tuition aid.

   Whether such a program violates the Establishment Clause,
the Court held, does not depend on whether the parent receiv-
ing tuition aid has a broader array of religious than secular
schools to choose from. This is because “[t]he constitutional-
ity of a neutral educational aid program simply does not turn
on whether and why, in a particular area, at a particular time,
most private schools are run by religious organizations, or
most recipients choose to use the aid at a religious school.” 

(emphasis added). The majority in Zelman made clear,
however, that a “neutral educational aid program” — or, as
the Court also put it, a “program of true private choice” — is
one that grants access to benefits without regard to religion.

, 662.

   The importance of providing equal access to benefits is
emphasized throughout Zelman. A common thread running
through indirect aid programs the Supreme Court has upheld
against Establishment Clause challenges, the Zelman Court
observed, is that they have been “neutral with respect to reli-
gion[ ] and provide[d] assistance directly to a broad class of
citizens.” 

(emphases added); see also 

; Witters v. Wash. Dep’t of Servs. for the Blind,

, 487 (1986); Zobrest v. Catalina Foothills Sch.
                 WINN v. ARIZONA CHRISTIAN SCHOOL                    14709
Dist., 

(1993).5 Likewise, under the Ohio voucher
program, the Court stressed, “[p]rogram benefits are available
to participating families on neutral terms, with no reference
to religion.” 

(emphases added).

   By contrast, as plaintiffs allege the Arizona Department of
Revenue applies the statute, access to Section 1089-funded
scholarships is not “available . . . on neutral terms, with no
reference to religion.” Parents who are unwilling to send their
child to a religious school may be denied access to program
benefits because, as plaintiffs allege, there are not a sufficient
number of scholarships available for use at secular schools.
Accordingly, these parents are shut out of the program alto-
gether, and at the very least their chances of receiving benefits
are harmed by their choice to send their child to a secular
school.

   This lack of access on a religiously neutral basis explains
why Section 1089 as operated by the Arizona Department of
Revenue would violate the Establishment Clause. This con-
clusion is entirely consistent with — and required by — the
Court’s analysis in Zelman. It is true that the majority in Zel-
man rejected Justice Souter’s view that the number of reli-
gious and secular schools participating in the Ohio voucher
program was relevant to its constitutionality. But before the
Court addressed Justice Souter’s concerns, it first identified
several features of the Ohio program that made it one of “true
private choice . . . , and thus constitutional.” 

features, the tuition aid distributed under the Ohio program
  5
    The program upheld in Mueller, Zelman explained, provided aid to
“ ‘all parents’ ” to pay for certain educational expenses at secular or reli-
gious schools. 

(quoting 

). Under the program in Witters, Zelman continued, “recipients gener-
ally were empowered to direct the aid to schools or institutions of their
own choosing.” 

(emphases added). Likewise, the program
upheld in Zobrest “ ‘distribute[d] benefits neutrally to any child qualifying
as disabled,’ ” without regard to religion. 

(quoting

).
14710         WINN v. ARIZONA CHRISTIAN SCHOOL
created “no ‘financial incentives’ that ‘skew[ed]’ the program
toward religious schools.” 

(quoting

). The Court recognized that
“[s]uch incentives ‘are not present where the aid is allocated
on the basis of neutral, secular criteria that neither favor nor
disfavor religion, and is made available to both religious and
secular beneficiaries on a nondiscriminatory basis.’ ” 

(emphasis added, alteration and ellipses omitted)
(quoting Agostini v. Felton, 

, 231 (1997)).

   The Arizona Department of Revenue permits scholarships
funded under Section 1089 not to be “made available to both
religious and secular beneficiaries on a nondiscriminatory
basis.” Therefore, Zelman requires a closer look at whether
the program, as applied, creates “financial incentives that
skew the program toward religious schools.” 

(alter-
ations and internal quotation marks omitted). This is why it is
relevant that, by allowing tax credits for contributions to dis-
criminatory STOs, the Arizona Department of Revenue has
created an overwhelming disparity in the number of scholar-
ships exclusively available for use at religious schools com-
pared to the number available for use at secular schools. See

.

   Consistent with Section 1089’s parental choice provision,
see Ariz. Rev. Stat. Ann. § 43-1089 (2005), the Arizona
Department of Revenue could apply the program to require
that STOs make state-funded scholarships “available to both
religious and secular beneficiaries on a nondiscriminatory
basis.” 

-54 (internal quotation marks
omitted). If Section 1089 were applied in this neutral manner,
data concerning the number of scholarships applied toward
religious schools versus secular schools would indeed be irrel-
evant to the Establishment Clause inquiry as long as the State
of Arizona otherwise provided students “a range of [secular]
educational choices.” 

                 WINN v. ARIZONA CHRISTIAN SCHOOL                    14711
   As Section 1089 is currently applied, however, the program
allows state-funded scholarships to be restricted to use at reli-
gious schools. It is therefore necessary to consider whether, in
reality, the program creates incentives for parents to send their
children to religious schools in order to gain access to bene-
fits. This conclusion is required by — and, at the very least,
consistent with — Zelman.6

                                    2.

   The dissent contends that the Arizona program is valid
because the State of Arizona has other programs in place that
provide secular educational options for those unable to obtain
a program scholarship. See 

(“The
Establishment Clause question is whether Ohio is coercing
parents into sending their children to religious schools, and
that question must be answered by evaluating all options Ohio
provides Cleveland schoolchildren, only one of which is to
obtain a program scholarship and then choose a religious
school.” (emphases added)). The dissent fails to recognize
Zelman’s holding that a program of “true private choice” is
one in which “[p]rogram benefits are available to participating
families on neutral terms, with no reference to religion.” 

. A short hypothetical will show why the dissent’s read-
ing of Zelman is untenable. Consider a program, instituted by
a state that provides an array of secular educational options,
that offers tax deductions exclusively to parents sending their
children to private schools. Each eligible parent receives a tax
  6
    The dissent’s hypothetical, about a world in which Section 1089 oper-
ates to restrict scholarships to use at secular schools, is premised on the
misunderstanding that, under the Free Exercise Clause, governments have
an affirmative obligation to fund religious educational options if they
decide to make secular options available. See Dissent 14725-26. But see
Locke v. Davey, 

(2004) (rejecting Free Exercise challenge
to statute providing postsecondary education scholarships but prohibiting
use of the scholarships for a degree in devotional theology from a religious
institution). Constitutional limitations on support for religion do not pre-
cisely mirror limitations on failures to support religion.
14712            WINN v. ARIZONA CHRISTIAN SCHOOL
deduction unrelated to the amount spent on tuition, thus
ensuring a windfall to parents who send their children to reli-
gious schools, which typically charge lower rates than secular
private schools. Assume this hypothetical program has a valid
secular purpose. For most potential recipients, however, bene-
fits under the program are, as a practical matter, available
only if the recipient chooses to send her child to a relatively
low-cost religious school. Thus, the unmistakable effect of the
program is to create special incentives to send one’s child to
a sectarian school.

  Under the dissent’s reading of Zelman, this hypothetical
program would easily withstand an Establishment Clause
challenge. Anyone can participate in the program. Anyone
who participates receives identical tax benefits. Anyone can
apply the tax benefits toward a private school of his or her
choice. See Dissent 14722. Moreover, for those who choose
not to participate in the program, the state provides an array
of public schooling options. See 

   This program is not a hypothetical; it is the New York tax
deduction scheme invalidated in Committee for Public Educa-
tion & Religious Liberty v. Nyquist, 

(1973) —
a case that Zelman distinguished but declined to overturn.7 See

(“In its attempt to enhance the
opportunities of the poor to choose between public and non-
  7
    The New York program also had provisions offering private school
tuition reimbursements, paid directly to the school, for parents who fell
below a certain income level and grants to private schools for the mainte-
nance and repair of their facilities. The Court treated each provision of the
program as severable and held that each provision separately violated the
Establishment Clause. See 

. Nothing in Zelman
suggests the Court would have upheld the tax deduction provision of the
New York program if it had been considered in isolation. See 

(observing that New York program “gave a package of bene-
fits exclusively to private schools and the parents of private school enroll-
ees” and “provided tax benefits ‘unrelated to the amount of money
actually expended by any parent on tuition,’ ensuring a windfall to parents
of children in religious schools” (quoting 

).
               WINN v. ARIZONA CHRISTIAN SCHOOL             14713
public education, the State has taken a step which can only be
regarded as one ‘advancing’ religion.”); 

(observing that the “ ‘function’ ” of the New York pro-
gram “was ‘unmistakably to provide desired financial support
for nonpublic, sectarian institutions’ ” (quoting 

(emphasis added))). Zelman explained that
Nyquist had “expressly reserved judgment with respect to ‘a
case involving some form of public assistance (e.g., scholar-
ships) made available generally without regard to the
sectarian-nonsectarian, or public-nonpublic nature of the insti-
tution benefitted.’ ” 

413 U.S. at 782-83).
The Ohio voucher program fit this description. The Arizona
Department of Revenue’s application of Section 1089 does
not.

   In Zelman, the Court clarified that “Nyquist does not gov-
ern neutral educational assistance programs that . . . offer aid
directly to a broad class of individual recipients defined with-
out regard to religion.” 

According to the dissent,
Section 1089 is such a “neutral educational assistance pro-
gram.” By the dissent’s logic, however, the New York pro-
gram invalidated in Nyquist would also be such a program.
Even more troublesome, a program that provided tax deduc-
tions exclusively to parents sending their children to religious
schools would also constitute a “neutral educational assis-
tance program” as long as the state had other programs in
place that provided secular educational options. It is this
result, not the outcome in Winn, “that simply cannot be recon-
ciled with Zelman.” Dissent 14724. The Supreme Court
elected not to overturn Nyquist, and we may not do so on its
behalf.

                     B.   Taxpayer Choice

   Next, the dissent challenges the panel’s conclusion that the
choices provided to taxpayers under Section 1089 — choices
that, plaintiffs allege, restrict parents’ access to secular educa-
tional scholarships — fail to render the program constitu-
14714         WINN v. ARIZONA CHRISTIAN SCHOOL
tional. See 

. Specifically, the
dissent contends the panel misapplied Larkin v. Grendel’s
Den, Inc., 

(1982). A careful reading of Winn
shows that the dissent exaggerates the panel’s reliance on
Larkin.

   Larkin does not control this case. See 

(emphasizing that “the delegation of scholarship funding
to individual taxpayers, such as in Section 1089, does less to
promote religion than the delegation of zoning authority to
churches,” such as provided under the statute at issue in
Larkin). Rather, the panel observed that Larkin’s holding

    illustrates that when a statute delegates “a power
    ordinarily vested in agencies of government” to a
    private party, see [Larkin, 459 U.S.] at 522, without
    reasonable assurance that the party’s choices will
    advance the secular purposes of the statute, any
    ensuing “perceived endorsement of a religious mes-
    sage” may be “reasonably attribut[ed]” to the gov-
    ernment.

-21 (second quotation from 

).

   Contrary to the dissent’s suggestion, Larkin’s holding is not
limited to cases where the state vests governmental powers in
a “pervasively sectarian organization.” Dissent 14735. In
Board of Education of Kiryas Joel Village School District v.
Grumet, 

(1994), for example, the Court applied
Larkin when considering an Establishment Clause challenge
to a statute creating a school district coextensive with a reli-
gious community. This statute vested power in individual tax-
payers, not in a religious organization, but the Court
nonetheless invalidated it. See 

(“The Establishment
Clause problem [at issue] is more subtle [than that in Larkin],
but it resembles the issue raised in Larkin to the extent that
the earlier case teaches that a State may not delegate its civic
                 WINN v. ARIZONA CHRISTIAN SCHOOL                    14715
authority to a group chosen according to a religious criteri-
on.”). Again, Winn does not suggest that the choices delegated
to taxpayers under the Arizona Department of Revenue’s con-
struction of Section 1089 are the constitutional equivalent of
the legislative action at issue in Kiryas Joel. That case makes
clear, however, that the delegation concerns identified in
Larkin are relevant to whether a reasonable, informed
observer would conclude that the choices delegated under
Section 1089 have the effect of promoting, or hindering, the
program’s secular purpose.8

   The dissent further argues that in terms of constraining par-
ents’ access to secular educational options, taxpayers’ choices
under Section 1089 are no more constitutionally problematic
than the choice of public schools not to participate in the Ohio
voucher program in Zelman. Dissent 14736. This misses the
point. Under the Ohio program, it was not the public schools’
choices that ensured “the Establishment Clause was not impli-
cated” — it was the parents’ choices. 

. Each parent had an equal choice under the Ohio program
as to whether to apply a tuition voucher toward a private
school, and the choice of one parent did not directly alter the
array of options available to another parent. Accordingly,
under the Ohio program, a parent’s choice as to how best to
educate her child had no coercive effect on another parent’s
choice whether to send her child to a secular or religious
school. The Ohio program thus provided each eligible parent
with a “genuine and independent private choice” whether to
direct assistance to a religious school. 

.
The choices given to parents under the program therefore
ensured “the circuit between government and religion was
broken.” 

public school’s decision not to accept
  8
    The dissent asserts that the “allocation of scholarship funds” is not a
traditional governmental function. Dissent 14735 n.20. This framing of the
question is unhelpfully narrow. Educational policy is certainly a traditional
government function, and the state’s decision to reimburse contributions
to private scholarship funds is indisputably an educational policy decision.
14716         WINN v. ARIZONA CHRISTIAN SCHOOL
a voucher could indeed frustrate the secular purpose of the
Ohio program, the Supreme Court did not rely on those
choices to conclude the program was constitutional.

   By contrast, the appellees in Winn rely on taxpayers’
choices in arguing that the current construction of Section
1089 is valid under the Establishment Clause. The effect of
these taxpayer choices, however, may be to harm the ability
of aspiring scholarship recipients to obtain a scholarship
available for use at a secular school. Accordingly, Section
1089 “delegat[es] to taxpayers a choice that, from the per-
spective of the program’s aid recipients, ‘deliberately skew[s]
incentives toward religious schools.’ ” 

(quoting 

)). Such choices are not, in
themselves, sufficient to render an educational aid program
valid under the Establishment Clause.

   In summary, it was crucial to Zelman’s holding that the
Ohio program afforded aid recipients a “genuine and indepen-
dent private choice” whether to direct the assistance they
received toward a religious school. 

.
Although the Arizona Department of Revenue’s application
of Section 1089 does not afford such choice, Winn carefully
determined that the choice provided to taxpayers is insuffi-
cient to ensure “the circuit between government and religion
was broken.” 

carefully laid out in the decision,
our court was correct to decline en banc review.

                   III.   Secular Purpose

   The dissent also faults Winn for concluding that Section
1089 may lack a valid secular purpose. 

-
12. This criticism is premature. The question before the panel
was “not whether Section 1089 in fact has a genuine, secular
purpose, but whether plaintiffs could prove, on the facts
alleged in the complaint, that it does not.” 

  As the dissent states, a “legislature’s stated reasons” for
enacting a statute “will generally get deference,” and must be
              WINN v. ARIZONA CHRISTIAN SCHOOL            14717
accepted as true except in “unusual cases where the claim was
an apparent sham, or the secular purpose secondary.” Mc-
Creary County v. ACLU of Ky., 

, 865 (2005).
The inquiry into whether a statute’s ostensible purpose is a
sham or secondary to a religious objective, however, must be
“undertaken from the perspective of ‘an objective observer,
one who takes account of the traditional external signs that
show up in the text, legislative history, and implementation of
the statute, or comparable official act.’ ” 

(quoting 

. The dissent’s posi-
tion would foreclose such an inquiry.

   The dissent surmises that in enacting Section 1089, “[t]he
legislature could hardly have had the ‘purpose’ of endorsing
religion when it set up a plan that, for all it knew, could have
resulted in absolutely no funding for religious entities.” Dis-
sent 14739 (emphasis added). The dissent makes this claim
without citing any evidence concerning what the legislature
actually knew about how Section 1089 would likely operate.
This is just as well, because no such evidence yet appears in
the record. But the dissent suggests it should not matter to us
whether the legislature knew that Section 1089 would result
in disproportionate funding being made available only for use
at religious institutions. The Supreme Court has cautioned us,
however, against evaluating a program’s purpose from the
perspective of an “absentminded objective observer, not one
presumed to be familiar with the history of the government’s
actions and competent to learn what history has to show.”

. In short, the dissent appears to
call for either a heightened pleading standard for Establish-
ment Clause claims — under which the plaintiff must allege
specific facts establishing that the legislature acted for an
invalid purpose — or an approach to evaluating a program’s
purpose that “would cut context out of the enquiry, to the
point of ignoring history.” 

   Regarding evidence of Section 1089’s implementation, the
dissent contends that because taxpayer contributions to STOs
14718            WINN v. ARIZONA CHRISTIAN SCHOOL
are private conduct, they are irrelevant to whether Section
1089 has a valid secular purpose. Notably, in an as-applied
context, at least two circuits have considered private conduct
under a government program to be probative of the program’s
purpose,9 and Winn does not foreclose the possibility that
plaintiffs can point to private conduct probative of whether
Section 1089 has a secular purpose. More probative, however,
is the government conduct on the part of the State of Arizona
in implementing Section 1089.

    The Arizona Department of Revenue implements Section
1089 by allowing individuals to claim tax credits for contribu-
tions to private STOs. See 

(“The
Supreme Court has recognized . . . that state tax policies such
as tax deductions, tax exemptions and tax credits are means
of ‘channeling . . . [state] assistance’ to private organizations
. . . .” (emphasis added, ellipses in original) (quoting 

)). According to plaintiffs’ complaint, the Ari-
zona Department of Revenue allows tax credits for contribu-
tions to STOs that provide scholarships only to religious
schools. Individuals’ contributions to STOs that discriminate
on the basis of religion are not, of themselves, probative of
Section 1089’s purpose. The fact that the Arizona Department
   9
     See Staley v. Harris County, Tex., 

, 513 (5th Cir. 2006);
Bonham v. D.C. Library Admin., 

, 1244-45 (D.C. Cir.
1993). In Staley, the Fifth Circuit considered the community response to
a government-sponsored monument of a local citizen carrying a Bible in
evaluating whether the monument had a valid secular purpose. 

(“[T]he fact that the monument, with the Bible, stood without
complaint [from citizens] for thirty-two years, supports the notion that the
original purpose was not objectively seen as predominantly religious.”). In
Bonham, Judge Mikva, reversing the district court’s dismissal under Rule
12(b)(6) of a pro se plaintiff’s claim that the closing of a public library on
Easter Sunday violated the Establishment Clause, observed: “In determin-
ing the legislative purpose of a law or government practice, courts gener-
ally look to . . . testimony of parties who participated in the enactment or
implementation of the challenged law or practice, historical context, and
the sequence of events leading to the passage of the law or the initiation
of the 

(emphasis added).
              WINN v. ARIZONA CHRISTIAN SCHOOL           14719
of Revenue gives tax credits for these contributions, however,
could be probative of legislative expectations as to how state
assistance under Section 1089 would be directed in practice.

   Accordingly, by declining to rehear this case en banc, we
appropriately rejected the suggestion that we should turn a
blind eye to the history and implementation of Section 1089
simply because the statute is facially neutral.

                      IV.   Conclusion

   This case required the panel to apply Zelman to an educa-
tional aid program that, according to the allegations of plain-
tiffs’ complaint, lets taxpayers choose to make state-
reimbursed contributions to private scholarship organizations,
but allows the organizations to restrict access to state-funded
scholarships on the basis of religion. Winn correctly held that
such a program would not, under Zelman, provide parents
with “genuine and independent private choice,” and that the
choice given to taxpayers under such a program could not be
treated as the constitutional equivalent of the choice given to
parents under the Ohio voucher program. Unlike the program
in Zelman, the program alleged here neither makes scholar-
ships available to parents on a religiously neutral basis nor
gives them a true private choice as to where to utilize the
scholarships. The panel correctly held that these allegations,
if proven true, could establish an Establishment Clause viola-
tion. We therefore concur in the denial of rehearing en banc.



O’SCANNLAIN, Circuit Judge, dissenting from the denial of
rehearing en banc, joined by KOZINSKI, Chief Judge,
KLEINFELD, GOULD, TALLMAN, BYBEE, BEA, and
N.R. SMITH, Circuit Judges:

  This case involves an Establishment Clause challenge to an
Arizona educational tax credit program that provides scholar-
14720            WINN v. ARIZONA CHRISTIAN SCHOOL
ships to students wishing to attend private schools. This case
is more notable, however, for what it does not involve: state
action advancing religion. The government does not direct
any aid to any religious school. Nor does the government
encourage, promote, or otherwise incentivize private actors to
direct aid to religious schools. Rather, “state aid reaches reli-
gious schools solely as a result of the numerous independent
decisions of private individuals.” Zelman v. Simmons-Harris,

, 655 (2002).

   Unable to find any forbidden state action, the district court
correctly dismissed the case on the pleadings. Sadly, our
three-judge panel reversed. See Winn v. Ariz. Christian Sch.
Tuition Org., 

(9th Cir. 2009). Because a pro-
gram of scrupulous “governmental neutrality between religion
and religion, and between religion and nonreligion,” Epper-
son v. Arkansas, 

, 104 (1968), cannot violate the
Establishment Clause, I respectfully dissent from our full
court’s regrettable denial of rehearing en banc.

   I dissent not only because Winn cannot be squared with the
Supreme Court’s mandate in Zelman, but also because the
panel’s holding casts a pall over comparable educational tax-
credit schemes in states across the nation and could derail leg-
islative efforts in four states within our circuit to create simi-
lar programs.1 In short, the panel’s conclusion invalidates an
increasingly popular method for providing school choice,
jeopardizing the educational opportunities of hundreds of
thousands of children nationwide.2
  1
     See Fla. Stat. § 220.187; Ga. Code Ann. § 48-7-29.16; Ind. Code § 6-
3.1-30.5; Iowa Code § 422.11S; 24 Pa. Stat. Ann. 20-2005-B; R.I. Gen.
Laws § 44-62-2; A.B. 279, 2009-10 Leg., Reg. Sess., § 1 (Ca. 2009); S.B.
342, 61st Leg., Reg. Sess., § 1(3)(b)-(c) (Mont. 2009); S.B. 289, 75th
Leg., Reg. Sess., § 6(1) (Nev. 2009); H.B. 2754, 75th Leg., Reg. Sess. (Or.
2009).
   2
     Such programs have operated without incident, perhaps because no one
has thought to challenge them post-Zelman. Cf., e.g., Bush v. Holmes, 

, 399 (Fla. 2006) (explaining that plaintiffs voluntarily dis-
missed a challenge to a Florida school choice program after Zelman).
                 WINN v. ARIZONA CHRISTIAN SCHOOL                    14721
                                     I

   Arizona law (“Section 1089”) allows individuals voluntar-
ily to contribute money to private, nonprofit corporations
known as “student tuition organizations” (“STOs”). Ariz.
Rev. Stat. Ann. § 43-1089(A). Anyone can form an STO, and
there are no constraints on a taxpayer’s ability to donate to an
STO of his choice. Should a taxpayer elect to direct funds to
an STO, that contribution is refunded via tax credits of up to
$500 for individual taxpayers and up to $1000 for married
couples filing jointly.3 

these funds to provide scholarships and tuition
grants to students attending schools within the state. 

1089(G). While essentially any private school is statutorily
eligible to receive scholarship monies,4 STOs may choose
which institutions they will support, so long as they provide
funds to more than one school. Id.5 Parents then decide which
  3
     The distinction the panel tries to draw between a tax credit and a
deduction, 

, can have no constitutional signifi-
cance. Both result in a reduction of the money paid by the taxpayer to the
government, with the amount of the reduction going to the designated
STO. The only practical difference is that with a deduction the taxpayer
must make a co-payment of his own, whereas with a credit there is no co-
payment. Of course, this favors richer taxpayers over poorer ones, as the
former are more able to afford a personal contribution. Moreover, in a pro-
gressive tax system, deductions most favor the taxpayers with the greatest
income. Not only does the value of the deduction increase with the taxpay-
er’s marginal rate, but so does the amount of government revenue that is
diverted at the taxpayer’s behest. It is difficult to see why such a regres-
sive regime (deductions) is constitutionally superior to the egalitarian tax
credit. Cf. Lochner v. New York, 

, 75 (1905) (Holmes, J., dis-
senting) (The Constitution “does not enact Mr. Herbert Spencer’s social
statics.”).
   4
     Schools that “discriminate on the basis of race, color, handicap, famil-
ial status or national origin” are ineligible. 

1089(G)(2).
   5
     Like virtually every other tax credit system, see supra note 1, the Ari-
zona statute requires STOs to provide scholarships “without limiting avail-
ability to only students of one school.” Ariz. Rev. Stat. Ann. § 43-
14722            WINN v. ARIZONA CHRISTIAN SCHOOL
private school they would like their child to attend, and apply
for scholarships from appropriate STOs.

   In sum, the state’s involvement stops with authorizing the
creation of STOs and making tax credits available. After that,
the government takes its hands off the wheel. Anyone can
create an STO. Anyone can contribute to any STO and
receive identical tax benefits. Anyone can apply for any
scholarship offered by any STO.

   Shortly after Section 1089’s enactment, the Arizona
Supreme Court held that the statute, on its face, did not vio-
late the Establishment Clause. See Kotterman v. Killian, 

(Ariz. 1999). Taxpayer plaintiffs then brought this
federal action, which was dismissed by the district court under
the Tax Injunction Act. See Winn v. Killian, 

,
1013 (9th Cir. 2002). After the suit was reinstated, see 

see also Hibbs v. Winn, 

(2004) (affirming
our opinion reversing its dismissal), the district court again
dismissed the action, this time on federal constitutional
grounds, see Winn v. Hibbs, 

(D. Ariz.
2005).

   Plaintiffs appealed. They allege (and no one disputes) that
in practice, some STOs make their scholarships available only
to students willing to attend religiously affiliated schools.

. While the majority of STOs do not
so limit their scholarships,6 plaintiffs maintain that those that
do receive the overwhelming majority of taxpayer contribu-
tions. See 

assert that the pool of avail-

1089(G)(3). And again, like most other schemes, see supra note 1, the
statute says STOs should allow children “to attend any qualified school of
their parents’ choice.” 

the model of clarity, this language
has been interpreted to mean STOs satisfy the statue by providing scholar-
ships to at least two schools.
   6
     Twenty-five of the fifty-five existing STOs limit scholarship awards to
religious schools.
                 WINN v. ARIZONA CHRISTIAN SCHOOL                   14723
able scholarship money is diminished for parents wishing to
send their children to secular schools. See 

tend that this disparity means Section 1089, as applied, imper-
missibly favors religion over nonreligion. See 

panel agreed and reversed the district
court’s dismissal, holding that “if plaintiffs’ allegations are
accepted as true, Section 1089 violates the Establishment
Clause.” See 

Concluding that the nature of the tax
credit made taxpayer contributions tantamount to government
funds, the panel found that Section 1089 potentially violated
both the purpose and effects prongs of Lemon v. Kurtzman,

(1971). See 

The fact that taxpay-
ers directed the majority of available funds to religious
schools, the panel reasoned, deprived parents of a “genuinely
independent and private choice[ ]” to send their children to
secular private schools. 

(internal quotation marks
and citation omitted). Accordingly, Section 1089 was not a
“neutral program of private choice and a reasonable observer
could . . . conclude that the aid reaching religious schools . . .
carries with it the imprimatur of government endorsement.”

-14 (internal quotation marks and citation omitted).7

                                    II

   I have no bone to pick with the manner in which the panel
frames the basic constitutional inquiry. We all understand that
the Establishment Clause “prevents a State from enacting
laws that have the ‘purpose’ or ‘effect’ of advancing or inhib-
iting religion.” 

. More often than
  7
    Make no mistake about the procedural posture of this decision. True,
the case will be remanded to the district court. But the panel holds that,
“if plaintiffs’ allegations are accepted as true, Section 1089 violates the
Establishment Clause.” 

. So far as I can tell, no one
disputes plaintiffs’ factual allegations about how the program operates in
practice. Thus, the panel leaves the district court with no choice but to
declare the program unconstitutional as applied, rendering the remand lit-
tle more than an empty formality.
14724            WINN v. ARIZONA CHRISTIAN SCHOOL
not, the Court determines whether these commands have been
violated by asking whether a “reasonable observer,” who is
“aware of the history and context underlying a challenged
program,” would conclude that the state has “endorsed” reli-
gion. 

(internal quotation marks and citation omit-
ted).

   The panel’s heavy emphasis on Zelman is also warranted.
In that case, the Supreme Court upheld an Ohio school
voucher program that provided tuition aid to Cleveland fami-
lies on the basis of need. 

The vouchers were
distributed directly to parents, who could choose to use the
scholarship money at any participating private, community,
magnet, or public school. 

The Court ruled that
a “neutral program of private choice, where state aid reaches
religious schools solely as a result of the numerous indepen-
dent decisions of private individuals” does not violate the
Establishment Clause. 

   It is in the application of these standards, however, that the
three-judge panel lost the forest for the trees. In doing so, it
reached a result that simply cannot be reconciled with Zelman.8

                                   III

   The panel is correct that a law may not have the “forbidden
‘effect’ of advancing . . . religion.” 

What the panel
seems to neglect, however, is that “[f]or a law to have forbid-
den ‘effects’ under Lemon, it must be fair to say that the gov-
ernment itself has advanced religion through its own activities
and influence.” Corp. of the Presiding Bishop of the Church
of Jesus Christ of Latter-Day Saints v. Amos, 

,
337 (1987).9
  8
    As the panel focused primarily on “effects,” rather than “purpose,” I
address these two Lemon prongs out of order.
  9
    For example, the panel asserts that Arizona parents are presented with
a choice that “deliberately skew[s] incentives toward religious schools.”
                  WINN v. ARIZONA CHRISTIAN SCHOOL                     14725
   I must confess that I am at a loss to understand how a rea-
sonable observer—one fully informed about all matters
related to the program—could conclude that the “government
itself” has endorsed religion in this case. Multiple layers of
private, individual choice separate the state from any religious
entanglement: the “government itself” is at least four times
removed from any aid to religious organizations. First, an
individual or group of individuals must choose to create an
STO. Second, that STO must then decide to provide scholar-
ships to religious schools. Third, taxpayers have to contribute
to the STO in question. Finally, parents need to apply for a
scholarship for their student. In every respect and at every
level, these are purely private choices, not government policy.
Under such circumstances, “government cannot, or at least
cannot easily, grant special favors that might lead to a reli-
gious establishment.” 

-53 (internal
quotation marks and citation omitted). Only after passing
through choice piled upon choice do government funds reach
religious organizations. That is not government endorsement:
that is government nonchalance.10

See 

(quoting 

). The actual
quotation from Zelman, however, includes a key qualifier: it condemns
only programs where “the State deliberately skewed incentives toward
religious 

(emphasis added). Since only state
action can violate the Establishment Clause, the panel’s omission is tell-
ing.
   10
      The panel makes much of the fact that Zelman discussed aid flowing
“directly” to parents. I submit that the Supreme Court used such terminol-
ogy for two reasons. First, that was the case before the Court: it had no
reason to pontificate on systems involving additional levels of private
choice. Second, the language emphasized that the voucher program did not
involve constitutionally problematic “direct” aid to religious institutions.
   The panel turns this language into a rallying cry to suggest that by filter-
ing aid through multiple levels of private choice—rather than a single
level—the state endorses religion. But that makes no sense. How can
increasing the separation between state and religion result in heightened
government endorsement?
14726         WINN v. ARIZONA CHRISTIAN SCHOOL
   To illustrate my point, consider the following hypothetical.
Assume the exact statutory scheme embodied in Section
1089: anyone can create an STO, anyone can donate to an
STO, and STOs can limit their scholarships to particular types
of schools. Now imagine that only agnostics decide to create
STOs. Imagine further that every STO refuses to provide
tuition assistance to religious schools. In short, assume there
is absolutely no money available for parents who want to send
their children to a religious school. Would the parents be jus-
tified in accusing the government of depriving their children
of school funds? Of course not.

   The foregoing example plainly shows that in this case, any
“endorsement” of religion arising from the disbursement of
state funds to religious entities turns wholly and completely
on the independent, uncoerced choices of private individuals.
The system Arizona created could just as easily have resulted
in a total dearth of funding for religious organizations as
opposed to the surfeit allegedly available. This feast or famine
is utterly out of the state’s hands. It simply cannot be, as the
panel claims, that the “scholarship program . . . skews aid in
favor of religious schools.” 

(emphasis
added). The “program” does no such thing: any “skew[ing]”
that occurs takes place because of private, not government
action. It is axiomatic that such action cannot violate the
Establishment Clause.

                               A

   The panel however, believes that under this multi-tiered
system, choice is the culprit, not the savior. After all, plain-
tiffs allege that it is “the choice delegated to taxpayers” which
“channels a disproportionate amount of government aid to
sectarian STOs [that] limit their scholarships to use at reli-
gious schools.” 

panel maintains, focused on
parental choice. 

Here, however, that choice is pur-
ported impermissibly to be “constrained” by the decisions of
taxpayers and STOs. 

In other words, the choices
               WINN v. ARIZONA CHRISTIAN SCHOOL            14727
of others deprive parents of their own “independent and pri-
vate choice[ ].” 

(internal quotation marks and cita-
tion omitted). They might want to send their children to
secular private schools, but scholarships are not readily avail-
able for that purpose. Moreover, the panel claims the alleged
abundance of funds from religious STOs creates an incentive
for these parents to enroll their children in religious schools.

The panel therefore holds that Section 1089
“fails to provide genuine opportunities for . . . parents to
select secular educational options for their school-age chil-
dren.” 

(internal quotation marks and citation omit-
ted).

   I admit that the panel’s conclusion with respect to the pur-
ported lack of parental choice finds support in Zelman. The
problem is, that support comes from Justice Souter’s dissent,
not the opinion of the Court. Several aspects of the majority’s
reasoning in that case make the Winn panel’s conclusion
infirm.

                                1

   By focusing generally on the scope of parental choice, the
Winn panel, like the Zelman dissent, is barking up the wrong
tree. The question is not whether a parent’s choice is some-
how limited or constrained, the question is whether the gov-
ernment has somehow limited or constrained the choice.

   In Zelman, Justice Souter accused the majority of allowing
external factors to “influenc[e] choices in a way that aims the
money in a religious 

(Souter, J.,
dissenting). Of the fifty-six private schools that participated in
the Cleveland voucher program, he noted, forty-six were reli-
gious. 

mind, this lack of a “wide array of private
nonreligious options” suggested that any “choice” was not
genuine. See 

Rather, he believed parents’ deci-
sionmaking process was skewed by “the fact that too few non-
religious school desks are available and few but religious
14728             WINN v. ARIZONA CHRISTIAN SCHOOL
schools can afford to accept more than a handful of voucher
students.” 

“For the overwhelming number of chil-
dren in the voucher scheme,” he concluded, “the only alterna-
tive to the public schools is religious.” 

not swayed
by the fact that these constraints were unrelated to state
action: “a Hobson’s choice is not a choice, whatever the rea-
son for being Hobsonian.” 

Justice Souter would
have struck down the Ohio voucher program because parents’
choice was influenced by factors beyond their control.

   Obviously, Justice Souter’s position did not carry the day.
“That 46 of the 56 private schools now participating in the
program are religious schools,” the majority explained, “does
not condemn it as a violation of the Establishment Clause.”

(majority opinion). For one thing, the Court noted
that the imbalance was not a function of government action.
See 

Moreover, “[t]o attribute constitutional sig-
nificance” to the availability of secular options, “would lead
to the absurd result that a neutral school-choice program
might be permissible in . . . some states [with a high concen-
tration of secular schools], but not in other States [where reli-
gious schools are plentiful].” 

To avoid this
absurdity, the majority held that “[t]he constitutionality of a
neutral educational aid program simply does not turn on
whether and why, in a particular area, at a particular time,12
most private schools are run by religious organizations, or
  11
      Earlier, I listed several state programs jeopardized by the panel’s hold-
ing. See supra note 1. Under the panel’s reasoning, those schemes could
be constitutional if taxpayers decided to provide more funds to secular,
rather than religious STOs. An identical program might then be constitu-
tional in one state and unconstitutional in another.
   12
      The “at a particular time” reference is especially significant. As dis-
cussed above, nothing in Section 1089 precludes any Arizona taxpayer,
tomorrow, from suddenly deciding to fund exclusively secular STOs. See
supra pp. 14709-11. The Supreme Court has twice declined to strike down
laws on the basis of such moving targets. See 

;
Muller v. Allen, 

, 401 (1983).
                  WINN v. ARIZONA CHRISTIAN SCHOOL                     14729
most recipients choose to use the aid at a religious school.” 

.13

    I see no meaningful distinction between the situation in
Zelman and the facts of this case. Both cases involve alleged
“constraints” on access to a scarce secular resource—
“nonreligious [private] school desks.” In Zelman, only ten of
the participating schools were secular. 

Parents were
thus “constrained” by third-party decisions to fund religious,
rather than secular schools. Here, while thirty out of fifty-five
STOs offer scholarships to secular schools, the majority of
program funds are allegedly concentrated in religious STOs.
Parents are thus “constrained” by the decisions of some STOs
to limit their scholarships to religious institutions, and tax-
payer choices to direct their funds to those STOs. The key
point is that in neither Zelman nor the case at hand are the
purported “constraints” government-induced. There is simply
no constitutionally significant distinction between a system
where—for reasons unattributable to state action—money is
available, but there are a limited number of schools to receive
it, and a system where schools may be available, but there is
a limited amount of money to spend. Under either scenario,
as Justice Souter bemoaned, “[f]or the overwhelming number
of children in the [program], the only alternative to the public
schools is religious.” 

(Souter, J., dissenting).

   I can go on. In Zelman, voucher funds could be used at par-
ticipating public schools in districts adjacent to Cleveland. 

(majority opinion). However, no such school “elected
to participate.” 

Parental choice was therefore “con-
strained” by the decisions of out-of-district public school
  13
     The Court went on to explain that it is “irrelevant . . . to the constitu-
tionality” of a government aid program that “a vast majority of program
benefits went to religious schools.” 

. The panel
distinguishes this point, claiming that the Zelman did not involve a situa-
tion where parental choice was “constrained.” 

n.14. As demonstrated below, that is simply not the case. See infra pp.
14728-31.
14730             WINN v. ARIZONA CHRISTIAN SCHOOL
administrators. Similarly, Ohio did not require private secular
schools to accept vouchers: they chose to do so. See 

n.4. Citing overcrowding or a desire for independence from
government funds, these schools could just as easily have
decided to opt out of the program. Alternatively, they could
have, for whatever reason, decided to close up shop. In either
scenario, parents again would be left with a reduced “choice”
to send their children to private, secular schools. Did the Zel-
man Court strike down the Ohio program for impermissibly
“delegating” such decisions to school administrators? Was
parental choice held to be unduly “constrained”? Of course
not. Instead, the Court said that the availability of a private
secular education, “in a particular area, at a particular time,”
was irrelevant to the constitutional inquiry. See 

; supra pp. 14728.14

   Ultimately, the panel seems to assume that parents must
have the same access to “nonreligious [private] school desks”
as they do to religious private school desks. But that was cer-
tainly not the case in Zelman, and the Ohio voucher program
was upheld. Indeed, such result is unattainable in any program
where the government is neutral with respect to religion and
nonreligion. If the government takes the constitutionally
required hands-off approach, external factors will define the
playing field. Contrary to the panel’s conclusion, the constitu-
tional inquiry “simply does not turn” on whatever influence
these factors might exert on parents. 

.
   14
      The Zelman Court’s comment that the “preponderance of religiously
affiliated private schools certainly did not arise as a result of the [voucher]
program” is also 

. The Court stated that the
imbalance was “a phenomenon common to many American cities.” 

In other words, the disparity was caused not by government action,
but rather by private predilections. The same can be said about the exis-
tence of religiously affiliated STOs and the disproportionate share of tax-
payer contributions they receive. The concentration of funds in religious
entities—and the resulting “constraint” on parental choice—“certainly did
not arise as a result of” any state action, but rather as a consequence of pri-
vate decisions. See supra pp. 14724-26.
                 WINN v. ARIZONA CHRISTIAN SCHOOL                    14731
Again, provided there is “no evidence that the State deliber-
ately skewed incentives toward religious schools,” there is no
Establishment Clause violation. 

(emphasis added);
see supra pp. 14723-26. As the Arizona tax credit program is
just as much a program of “true private choice” as the pro-
gram in 

, the panel erred in reinstat-
ing the constitutional challenge.15

                                     2

   In rejecting Justice Souter’s position, the Zelman majority
also emphasized that he was asking the wrong question.
Rather than focusing narrowly on the challenged voucher pro-
gram, the majority explained that the “Establishment Clause
question is whether Ohio is coercing parents into sending
their children to religious schools, and that question must be
answered by evaluating all options Ohio provides Cleveland
schoolchildren, only one of which is to obtain a [voucher].”

-56. Because the Winn panel adopts Justice Souter’s
overly restrictive approach, rather than assessing “all options”
available to Arizona students, its result is similarly flawed.16

   Indeed, the panel overtly limited its parental-choice inquiry
to “the range of educational choices the STO-administered
scholarship programs offer.” 

. It “re-
ject[ed] the suggestion that the mere existence of the public
school system guarantees that any scholarship program pro-
vides for genuine private choice.” 

latter state-
ment may be true, it is also something of a non sequitur. No
one claims the existence of a public school system grants a
  15
      For this reason, the Winn panel’s reliance on Committee for Public
Education & Religious Liberty v. Nyquist, 

(1973), is mis-
placed. See 

-62.
   16
      Interestingly, the Supreme Court decided to italicize “all options” in
Zelman, and “government itself” in Amos. Maybe the justices thought
these requirements were important. 

; 

.
14732            WINN v. ARIZONA CHRISTIAN SCHOOL
state license to ignore the Establishment Clause. The question,
as Zelman instructs, is whether Arizona is “coercing parents
into sending their children to religious schools,” a question
which must be answered by evaluating “all options” Arizona
provides its 

.

   The panel did not even engage in this inquiry. Had it done
so, it would have discovered that Section 1089 is but one of
a “range of educational choices” available to parents of
school-aged children. 

; see also 

(noting that the “Arizona Legislature has, in recent
years, expanded the options available in public education”
and listing some of those options). Arizona’s public schools
must provide for open enrollment, allowing parents to send
their children, tuition-free, to schools of their choice. Ariz.
Rev. Stat. Ann. § 15-816.01(A). Tax credits are available for
donations to public schools for “extracurricular activities or
character education.” Ariz. Rev. Stat. Ann. § 43-1089.01. An
extensive system of charter schools “provide[s] additional
academic choices for parents and pupils.” 

Homeschooling is permitted and protected. 

802-03. Indeed, Section 1089 itself offers parents yet another
alternative: they can create their own STO and solicit dona-
tions for use at secular private schools. These alternative edu-
cational opportunities mirror those the Court took into
consideration in Zelman. 

(“Cleveland
schoolchildren enjoy a range of educational choices: They
may remain in public school as before, remain in public
school with publicly funded tutoring aid, obtain a scholarship
and choose a religious school, obtain a scholarship and choose
a nonreligious private school, enroll in a community school,
or enroll in a magnet school.”).18
  17
      Out of the 4,000 plus charters schools across the country, 478 are in
Arizona.       See       Arizona      Charter      Schools      Association,
http://www.azcharters.org/pages/schools-basic-statistics (last visited July
25, 2009).
   18
      As the district court observed, parents are actually discouraged from
sending their children to private religious schools. “An Arizona student
                  WINN v. ARIZONA CHRISTIAN SCHOOL                      14733
   This is no Hobson’s choice. Far from “coercing” parents
into sending their children to religious schools, Arizona pro-
vides a wide variety of secular alternatives. “Any objective
observer familiar with the full history and context of [Section
1089] would reasonably view it as one aspect of a broader
undertaking . . . .” 

By shutting its eyes to the host
of options available to Arizona parents, the panel’s opinion
directly conflicts with Zelman.19

                                      B

   As demonstrated by the foregoing arguments, the Arizona
program provides parents with “true private choice.” That
established, the panel’s discussion of taxpayer choice
becomes surplusage. Indeed, is its curious focus on “taxpayer
choice” an apt analogy at all? I suggest that Winn’s reliance
on Larkin v. Grendel’s Den, Inc., 

(1982), is
utterly mistaken.

   The thrust of the panel’s reasoning is that taxpayer choice

may attend any public school without cost . . . . In contrast, the average
scholarship paid by STOs in 2003 was $1,222, a sum unlikely to cover all
of the costs of private school attendance.” 

(citations and footnote omitted); see also 

(“Families . . . have a financial disincentive to choose a private religious
school over other schools. Parents that choose to . . . enroll their children
in a private school . . . must copay a portion of the school’s tuition. Fami-
lies that choose a community school, magnet school, or traditional public
school pay nothing. [This] clearly dispel[s] the claim that the program
creates financial incentives for parents to choose a sectarian school.”
(internal quotation marks, alterations, and citation omitted)).
   19
      The concurrence argues that this reading of Zelman is inconsistent
with Nyquist. Concurrence at 14712-13. The Court invalidated the New
York tax program at issue in Nyquist because its tuition reimbursements
were designed “ ‘explicitly to offer . . . an incentive to parents to send their
children to sectarian schools.’ ” 

(quoting 

). As I have explained above, the plethora of choices
available to Arizona parents demonstrates that Section 1089 has no effect
of incentivizing religious schools over sectarian schools.
14734          WINN v. ARIZONA CHRISTIAN SCHOOL
is not a valid substitute for the parental choice allegedly at the
core of Zelman. I am not certain, however, that parental
choice was as central to the reasoning of Zelman as the panel
would have it. While that opinion does repeatedly refer to aid
“recipients,” see 

(listing citations), at
other times, it refers only to private, nongovernmental choice,
see, e.g., 

(describing programs where
“government aid reaches religious schools only as a result of
the genuine and independent choices of private individuals”);

(stating that “no reasonable observer” would find
government endorsement where “state aid reaches religious
schools solely as a result of the numerous independent deci-
sions of private individuals”). Significantly, Zelman seems
most concerned about preventing the state from reaching out
to “grant special favors that might lead to a religious estab-
lishment.” 

-53 (internal quotation marks and citation
omitted). So long as “favors” are doled out independent of
state action, the Establishment Clause—which again, prohib-
its the “government itself” from endorsing religion—is not
offended.

   I further submit that under the endorsement test, any level
of attenuation between government action and aid to religion
necessarily reduces the likelihood that a “reasonable observ-
er” will find impermissible government approbation. There
can be no doubt that taxpayer choice contributes to that atten-
uation. Thus, the panel’s analysis of whether the choice Sec-
tion 1089 provides to taxpayers ensures that “ ‘the circuit
between government and religion was broken’ ” is beside the
point. 

(quoting 

). The self-evident fact is that by “delegating” the choice
to taxpayers, the government already broke the circuit.

  Nonetheless, the panel contends a reasonable observer
would consider two factors when deciding whether a program
of individual choice violates the Establishment Clause: the
“role the person making the choice occupies in the structure
of the program,” 

and “whether the choice dele-
                 WINN v. ARIZONA CHRISTIAN SCHOOL                    14735
gated . . . has the effect of promoting, or hindering, the pro-
gram’s secular purpose,” 

Regarding the former,
the panel determined there was “no ‘effective means of guar-
anteeing’ ” that taxpayers would exercise their choice “ ‘ex-
clusively for secular, neutral, and nonideological purposes.’ ”

(quoting 

). Parents, on the
other hand, have “incentives to apply the program’s aid based
on their children’s educational interests instead of on sectarian
considerations.” 

As for the latter, the panel con-
cluded that taxpayers thwarted the secular purpose of the stat-
ute insofar as their contributions narrowed the range of
available educational alternatives. 

   One could see how a reasonable observer in Larkin could
perceive government endorsement of religion from the “role
the [entity] making the choice” played in the scheme. Maybe
I am stating the obvious, but a large part of that perception
might rest on the fact that in Larkin, the state delegated legis-
lative authority—the ability to veto liquor licenses—to
churches. 

. I say again: churches. Under
such circumstances it is completely unsurprising that a rea-
sonable observer would conclude that this “joint exercise of
legislative authority by Church and State provides a signifi-
cant symbolic benefit to religion in the minds of some.” 

To what pervasively sectarian organization has
Arizona “delegated” the choice at issue in this case? The Ari-
zona taxpayer.20 When perceived endorsement of religion is at
issue, state cooperation with churches is a far cry from state
cooperation with taxpayers.21
  20
      Additionally, the authority delegated in Larkin was absolute veto
power in an area of traditional government 

.
Here, each individual taxpayer exercises only a modicum of control over
the allocation of scholarship funds.
   21
      Board of Education of Kiryas Joel Village School District v. Grumet,

(1994), does not support the panel’s analysis. In Larkin, leg-
islative authority was delegated to churches. In Kiryas Joel, the state cre-
ated a school district such that a particular religious group would have
“exclusive control of the political 

. Both
actions displayed overt religious bias. “Delegation” to the Arizona tax-
payer does not.
14736         WINN v. ARIZONA CHRISTIAN SCHOOL
   Moreover, I disagree with the panel’s conclusion that par-
ents are somehow less motivated to promote religious objec-
tives than taxpayers generally. As anyone who has grown up
in a religious household will tell you, schooling decisions are
as frequently made on the basis of religious considerations as
they are on purely secular academic grounds. At the very
least, sectarian considerations factor into the equation of what
is in the child’s best interests educationally. Thus, whether it
resides with the taxpayer or the parent, once the choice is
made available, the state has no “effective means of guaran-
teeing” that it will be exercised “exclusively for secular, neu-
tral, and nonideological purposes.” 

(internal
quotation marks and citation omitted). By contrast with
engaging in pseudo-psychological inquires into motivation,
under Zelman, we need only satisfy ourselves that the choice,
whatever it is, is made by a private actor, not by the govern-
ment.

   With respect to taxpayers’ ability to “thwart” the secular
purpose of the statute, as discussed above, actors in any pro-
gram of true private choice will have this ability. See supra
pp. 14728-30. In Zelman, for example, the purpose of provid-
ing a broad range of educational opportunities was “thwarted”
by the decisions of neighboring public-school administrators
to decline program vouchers. See supra pp. 14729-30. The
goal could be similarly “thwarted” if secular private school
administrators decided to pull out of the program. See supra
pp. 14729-30—. An inherent reality of true private choice
programs cannot condemn Section 1089.

  Ultimately, the panel appears to argue that Arizona’s
scheme is flawed because it essentially delegates to a private
entity something the state could not constitutionally achieve
by the exercise of its own powers; here, the promotion of reli-
gious education. See 

; see also 

(citing Norwood v. Harrison, 

, 465 (1973)).
That may well be true, but as the panel’s own citation indi-
cates, for that to be the case, the state must somehow “induce,
                  WINN v. ARIZONA CHRISTIAN SCHOOL                     14737
encourage or promote private persons to accomplish what it
is constitutionally forbidden to accomplish.” 

(emphases added). At the risk of beating a dead
horse, I repeat that the state here has done nothing to cajole
parents, STOs, or taxpayers into supporting religious educa-
tion. The state has simply said, if you donate to the STO of
your choice, you get a tax credit. Such action in no way
induces, encourages, or promotes private parties to aid reli-
gion.22

                                     IV

   The panel also holds that plaintiffs have alleged facts sug-
gesting Section 1089 was not “enacted for . . . [a] valid secu-
lar purpose.” 

(internal quotation
marks and citation omitted). The panel reaches this conclu-
sion despite conceding that the statute is facially neutral with
respect to religion. See 

Nothing in the legisla-
tive history suggests that the driving force behind the bill was
anything other than the desire to provide “equal access to a
wide range of schooling options for students of every income
level.” Id.; see also 

(“A state’s deci-
sion to defray the cost of educational expenses incurred by
parents—regardless of the type of schools their children
attend—evidences a purpose that is both secular and under-
standable.”). Nonetheless, the panel maintains that plaintiffs
could prove, based on how Section 1089 operates in practice,
that this “secular and valid” purpose is a sham. 

-12.

   From its citation to McCreary County v. ACLU, 

     The concurrence asserts that “[t]he effect of these taxpayer choices . . .
may be to harm the ability of aspiring scholarship recipients to obtain a
scholarship available for use at a secular school.” Concurrence at 14716.
I fail to see how Section 1089—which permits tax deductions for gifts to
both religious and secular scholarship funds—harms a student’s ability to
obtain a scholarship to a secular school.
14738          WINN v. ARIZONA CHRISTIAN SCHOOL
844 (2005), the panel seems to argue that the very enactment
of Section 1089 “bespoke” a religious purpose. 

. But how can this be so? McCreary does say that
government action can be so “patently religious” that its non-
secular nature is 

. The examples pro-
vided, however, are situations where the state mandated Bible
study, the teaching of creationism, and prayer in schools. 

Setting up a tax credit program to provide scholar-
ships to children generally is hardly of the same ilk.

   To the extent the panel claims that the manner in which
Section 1089 has been implemented reveals the stated secular
purpose to be a sham, their arguments are similarly unpersua-
sive. First, the Supreme Court has recognized that a “legisla-
ture’s stated reasons will generally get deference,” deference
only abandoned in “those unusual cases where the claim was
an apparent sham.” 

Nothing in the plaintiffs
allegations suggest this is one of those “unusual cases,” and
as setting up a tax credit program is not a “patently religious”
act, there is nothing “apparent” about any purported sham.
Second, the implementation inquiry centers on actions taken
by the government. See 

(stating that the inquiry
turns on the “traditional external signs that show up in the
text, legislative history, and implementation of the statute, or
comparable official act”) (internal quotation marks and cita-
tions omitted); 

(questioning the government’s
newly proffered purposes after it altered a Ten Command-
ments display in an attempt to mitigate previously stated sec-
tarian purposes). Here, the alleged impropriety arises from
taxpayer, not government action. Third, the panel’s holding
turns on plaintiff’s allegation that “in practice STOs are per-
mitted to restrict the use of their scholarships to use at certain
religious schools.” 

. But that result is
apparent from the statute itself, which is satisfied so long as
STOs provide scholarships to two or more schools, see supra
note 5, a fact plaintiffs themselves recognize in their com-
                WINN v. ARIZONA CHRISTIAN SCHOOL                14739
plaint. That an STO may independently decide to limit its
scholarships does not make a religious purpose “apparent.”23

   Ultimately, the crux of the panel’s purpose holding turns on
matters previously discussed under the effects prong: a non-
secular purpose could be inferred from the fact that, at a given
moment, the bulk of scholarship money is available only for
use at religious schools. But as detailed above, money flows
to religious institutions entirely at the whim of nongovern-
mental actors: taxpayers or STOs. The legislature could
hardly have had the “purpose” of endorsing religion when it
set up a plan that, for all it knew, could have resulted in abso-
lutely no funding for religious entities. See supra pp.
14723-26. This moving target is irrelevant to the Establish-
ment Clause inquiry. See supra pp. 14727-31.

                                  V

   The layer upon layer of private choice built into this pro-
gram ensures that “the circuit between government and reli-
gion [is] broken.” 

. Try as it may, the
panel cannot complete such circuit. Ultimately, nothing in the
panel opinion grapples with the fact that Arizona does nothing
to encourage, to promote, or otherwise to incentivize private
actors to direct aid to religious schools. Nothing explains how
“the government itself has advanced religion through its own
activities and influence.” 

. Nothing
points to any “evidence that the State deliberately skewed
incentives toward religious schools.” 

(emphasis added). Nothing shows how Section 1089 enables
Arizona to “grant special favors that might lead to a religious
establishment.” 

-53 (internal quotation marks and
citation omitted).
  23
     Additionally, any inquiry into purpose must look at context. See

, 864, 866. As discussed above, Section 1089
was enacted amidst a broader effort to increase alternative educational
opportunities. See supra pp. 14730-33; see also 

.
14740            WINN v. ARIZONA CHRISTIAN SCHOOL
   But the three-judge panel can hardly be faulted for these
omissions: it cannot manufacture what does not exist.24 What
does exist is a tax credit system that relies entirely on private
choice. Individuals choose to create an STO. STOs choose to
limit their funds to certain schools. Taxpayers choose to
donate. Parents choose to apply for scholarships. In truth,
everyone in Arizona has a choice—everyone except the gov-
ernment. No reasonable observer would think this lengthy
chain of choice suggests the government has endorsed reli-
gion.

   Because the three-judge panel’s decision strays from estab-
lished Supreme Court precedent, and because it jeopardizes
the educational opportunities of thousands of children who
enjoy the benefits of Section 1089 and related programs
across the nation, I must respectfully dissent from our court’s
regrettable failure to rehear this case en banc.




  24
     Cf. Compassion in Dying v. Washington, 

, 1446-47 (9th
Cir. 1996) (Trott, J., dissenting from denial of rehearing en banc) (“No
magician—not David Copperfield, not even Harry Houdini—can produce
a rabbit from a hat unless the rabbit is in the hat to begin with. Moreover,
if a hat does not contain such an animal, a magician cannot claim that any-
thing he is able to produce from it is in fact a rabbit, no matter how sincere
he may be or how great his forensic skills. All of this has something to do
with basic physics.”).
                             PRINTED FOR
                   ADMINISTRATIVE OFFICE—U.S. COURTS
                BY THOMSON REUTERS/WEST—SAN FRANCISCO

The summary, which does not constitute a part of the opinion of the court, is copyrighted
                          © 2009 Thomson Reuters/West.