Opinion by Judge BERZON; Dissent by Judge FERNANDEZ; Dissent by Judge KLEINFELD.
BERZON, Circuit Judge:
The pivotal question in this case is whether a city may, using tax funds and public employees, provide stage lighting, sound, and other special-event services for a sectarian religious organization’s prayer service held in the bandshell of a public park. The City of Tucson believed that to do so would violate the Establishment Clause of the First Amendment. For that reason, although Tucson, through its Civic Events Fund (“Fund”), does provide such [1067]*1067support to certain events held in public parks, Tucson declined the National Day of Prayer Committee’s request that it provide the equipment and services for its National Day of Prayer gathering. We agree with Tucson that although the Establishment Clause remains a “blurred, indistinct and variable barrier” to government support for religious activity, Lemon v. Kurtzman, 403 U.S. 602, 614, 91 S.Ct. 2105 (1971), “the circumstances of [the] particular relationship,” id., between Tucson and recipients of Civic Events Fund support is such that Tucson was correct in concluding that the requested funding would have fallen on the Establishment Clause side of that “serpentine” wall. McCollum v. Board of Educ., 333 U.S. 203, 238, 68 S.Ct. 461, 92 L.Ed. 649 (1948) (Jackson, J., concurring). Tucson’s decision to refuse Civic Events Fund support to the prayer service’s organizers therefore did not run afoul of another First Amendment proscription, against abridging freedom of speech.
Background
I. The Tucson National Day of Prayer Gathering
The Tucson National Day of Prayer Committee (“Prayer Committee”) is an organization that, according to its funding application to the City, requires its members to “pledge [a] specific religious belief,” namely, Christian, and terminates the membership of any member who does not do so.1 In the spring of 1997, the Prayer Committee organized a National Day of Prayer gathering described in that application as “an annual gathering of Tucson Christians.” As planned and as carried out, the event, open to the public; involved “a time of prayer and worship” led by “[p]astors from nine different churches,” and two choirs. More than 500 area church congregations were invited to participate in prayers “for local, state and national issues.” Contributions were collected from attendees for the purpose of paying for this event and future similar events.
The Prayer Committee 'chose to hold its National Day of Prayer service in and around a bandshell in Tucson’s Reid Park. The organizers applied for and received a permit from the City to use the bandshell, and the event went forward as planned. Tucson in no way prevented the Prayer Committee from holding its prayer service in Reid Park nor interfered with its doing so, instead allowing the Prayer Committee’s organizers and members to use the park, as other citizens may do, “for purposes of assembly, communicating thoughts between citizens, and discussing public questions.” Hague v. Committee for Indus. Org., 307 U.S. 496, 515, 59 S.Ct. 954, 83 L.Ed. 1423 (1939).2
The Prayer Day’s organizers, including the appellants here, Patricia and Robert [1068]*1068Gentala, were not seeking, however, simply to use the park for their event. They also wanted the City to cover the costs of city-supplied lighting and audio systems. The Prayer Committee applied for free use of $840 worth of such equipment and related services for which the City ordinarily charges a fee, observing that “without help, we will have [an] inadequate stage and sound package, as previous years did.”
II. The Tucson Civic Events Fund
The City does fund such equipment and services for some private events held in the City’s parks, through a fund denominated the “Civic Events Fund.” Tucson, however, refused the Prayer Committee’s request for support, on the basis of an established policy precluding funding of events directly supporting religious organizations. The reason for the policy, as Tucson notified Patricia Gentala in denying her funding request and has consistently maintained in this litigation, is to avoid Establishment Clause violations.3
Modern Establishment Clause doctrine is a species of constitutional inquiry which, as we shall develop as we go on, abjures absolutist doctrines and emphasizes contextual analysis instead, and in which the devil is, consequently, emphatically in the details. We therefore focus here on the features of Tucson’s Civic Events Fund program with, perhaps, unusual specificity.
Tucson charges small fees ($75) for use of bandshells in city parks, but that general fee is not at issue in this case. Additionally, Tucson will provide, for specified fees, certain equipment and services for use at public facilities by any “civic, social, religious, charitable, commercial or other users.” The Prayer Committee is therefore eligible to use the available equipment as long as it pays the applicable fee.
The available rental equipment includes, for example, display booths, portable stages, bleachers, portable public address systems, microphones, speakers, extension cords, automated refuse containers, lighting systems and serving tables. For audio and lighting equipment used at the band-shell in Reid Park, the City requires that its staff operate the equipment. The rental fees charged for the use of that equipment therefore include labor charges, and City employees operate the equipment during the event. Ordinarily, an event’s organizers pay the necéssary fees to the pertinent city department, which, for lighting and audio equipment, is the Parks and Recreation Department.
The City has determined that, aside from allowing the public to use the parks, facilities such as bandshells, and city-owned equipment, the City will “support and encourage” certain “Civic Events.” The kinds of events that the City wishes affirmatively to support include those that: “celebrate and commemorate the historical, cultural and ethnic heritage of the City and the nation, or increase the community’s knowledge and understanding of critical issues ... [;] generate broad community appeal and participation[;] instill civic pride in the City, state, or nation[;] contribute to tourism[;] or are identified as unique community events.”
The City will recognize as a Civic Event only events sponsored by non-profit organizations or “individuals conducting the events on a non-profit basis,” that is, “with fund-raising proceeds used for a community benefit.” This limitation means that, whether or not they meet the substantive criteria for Civic Event designation, the City does not “support and encourage” corporate picnics, commercial concerts, or private birthday or anniversary celebrations in city parks (although the City does, as we understand the record, permit organizers of such- events to use city-owned equipment and personnel if they pay the required fees).4 Also excluded from Civic [1069]*1069Event designation and support are events sponsored by organizations that already receive city funding, whether directly or through rent waivers for use of city property. Finally, and critically here, among the events that “may [not] be considered for City'support through the Civic Event process” are “events held in direct support of religious organizations.”
As part of its support and encouragement of Civic Events, Tucson maintains a Civic Events Fund, consisting of moneys appropriated from its general coffers and derived from tax revenue, among other sources. Organizers of eligible events may apply for payment from the Fund of any fees incurred for use of city equipment or services. When such financial support is approved, however, the City (with limited exceptions) does not provide any money directly to the event’s sponsors. The city department that would ordinarily bill the event’s sponsors instead bills and receives payment from the Fund. Of course, both the Fund account and the departmental account are city accounts. To implement its subsidy program, therefore, the City simply transfers on its books the amount of any fees from the Fund to the appropriate department, as “a budgetary means of detailing the costs City departments incur providing in-kind support to Civic Events.”
Organizations receiving Fund support are required to: maintain financial records concerning the event and, on request, “provide City staff prompt access to such financial records and any additional financial information about the sponsoring organization”; obtain liability insurance for the event;5 and coordinate the planning of the event with a city Civic Event Coordinator. Additionally, to receive Civic Event Fund support, event organizers must promise to publicize the City’s contribution of services in their event advertising and in an announcement made during the event.
In the 1996-97 and 1997-98 fiscal years, the City appropriated $170,760 from its General Fund to the Civic Events Fund. A subcommittee of the City Council made decisions on applications for support of less of than $1,500; the Mayor and the full Council ruled on larger requests.
The Fund application and a document entitled “Civic Event Fund Policy Statement and Evaluation Criteria” spell out the criteria for funding in great detail. Among those criteria, in addition to the substantive ones summarized above, are: the need for support, including whether the sponsoring organization can pay some or all of the fees itself and whether it has explored other sources of support; whether the funding request is a first-time request or not (“The Mayor and Council believe[ ] that events should eventually become self-reliant.... To encourage events to become self-reliant, City support provided in subsequent years may be gradually reduced.”); and whether the event “generates broad community appeal and participation,” although “special consideration will ... be given to small first-time events conducted by organizations with limited resources.”
Although, as the detailed criteria summarized above indicate, these funding decisions are discretionary, in the only year for which there is complete information in the record there were sufficient funds available that most non-profit groups sponsoring eligible events received funding if they asked for it. In that single fiscal year, 1995, 23 events out of 26 that met the threshold criteria were funded; the Hu[1070]*1070mane Society-sponsored Fun Day in the Park, the Sister Cities Fund Raiser, and the Casa Car Show were not funded, for reasons that do not appear in the record.
The City has in the past implemented the exclusion for events directly supporting religious organizations by refusing to pay fees incurred by fundraisers for area religious schools and churches, although the events themselves were not uniformly religious in character.6 At the same time, the Fund supported other events sponsored by religiously-affiliated organizations when the events did not directly benefit the groups themselves,7 and events with religious themes sponsored by secular organizations. For example, in 1993, the Fund supported a Las Posadas pageant depicting theatrically events surrounding the birth of Jesus. The Las Posadas event was sponsored not by a religious group, however, but by the Tucson Festival Society and a local public school.
III. The Litigation
Two weeks before the Prayer Day event, Patricia Gentala, chair of the Prayer Committee, applied for $340 in support from the Fund for lighting and sound equipment and services. On the required application form, she identified the event as coming within the “Historical” category, enclosed a detailed description of the event, summarized above, and noted that there would be a “[f]ree will offer[ ]ing taken at close of program,” with any extra money from that collection to be applied to the following year’s event.
The Civic Events subcommittee denied Gentala’s application because the event was “in direct support of a religious organization.” Gentala sought City Council review of the subcommittee’s denial of the Prayer Committee’s application for support, and asked the City to reconsider its policy of categorically excluding “events in direct support of religious organizations” from consideration for Fund reimbursement. The City Council concluded that the denial was proper, explaining that it would violate the separation of church and state for the City to sponsor any event directly supporting a religious group.
A few weeks later, Patricia Gentala and her husband Robert Gentala, who also was actively involved in the Prayer Committee, filed this lawsuit, claiming that the exclusion of “events in direct support of religious organizations” from the Civic Events Fund violated the Free Speech, Free Exercise, and Establishment Clauses of the First Amendment,8 and seeking preliminary and permanent injunctions against excluding the National Day of Prayer and other religious groups from Civic Events Fund support. The Gentalas styled their complaint as both a facial challenge to the policy and a challenge to the policy as applied to them.
On September 30, 1997, the district court denied the Gentalas’ motions for an injunction. The district court recognized that the First Amendment’s speech protection “extends to religious speech and expression, including prayer and worship,” and that, where the government estab[1071]*1071lishes, by tradition or design, a public forum for communicative activity, exclusion of religious speech and expression can be justified only if necessitated by Establishment Clause considerations. The district court noted that the case on which the Gentalas most heavily rely is Rosenberger v. Rector & Visitors of the University of Virginia, 515 U.S. 819, 115 S.Ct. 2510, 132 L.Ed.2d 700 (1995), in which the Supreme Court held unconstitutional the exclusion of a religiously-oriented publication from a university program for funding the printing costs of student publications from mandatory student fees. Rosenberger, the district court recognized, is similar to this ease in some respects but dissimilar in others. In particular, the district court stressed, Rosenberger involved neither the expenditure of tax-derived funds nor funding for a religious organization, while in this case “Plaintiffs represent religious groups seeking tax dollars to support a religious event.” As a result, the district court concluded, Rosenberger is not determinative here, and “the dispositive inquiry in this case must be made under the Establishment Clause.”
Conducting that inquiry, the district court began by observing that, in cases raising issues at the intersection of the Free Speech and Establishment Clauses, “ ‘specific features of a particular government action,’ ” not any “ ‘Grand Unified Theory that may turn out to be neither grand nor unified,’ ” (quoting Rosenberger, 515 U.S. at 853, 115 S.Ct. 2510 (O’Connor, J., concurring)), are central in deciding whether the government is, on the one hand, required to support religiously-oriented speech activities or, on the other hand, precluded from doing so by the Establishment Clause. Noting that Rosen-berger had stressed that “if the State pays a church’s bills it is subsidizing it, and we must guard against this abuse,” 515 U.S. at 844, the district court held that compelling Establishment Clause considerations justified the City’s denial of the Prayer Committee’s funding request because: (1) the Prayer Committee was seeking access to general tax funds; (2) the Prayer Committee is a sectarian religious organization; and (3) if the event were “paid for, equipped by, and operated by the City,” there would be a distinct likelihood of creating “the impression that the City was engaged in a costly joint-enterprise with a religious group for religious purposes.”
- The district court also rejected the Gen-talas’ contention that the degree of discretion available to the risk manager with regard to the implementation of the insurance requirement constituted an impermissible prior restraint of the Gentalas’ speech.
The Gentalas appealed. We affirm, for reasons in most respects similar to those relied upon by the district court.
Standard of Review
Observing that the critical facts were undisputed, the district court found that the Gentalas had suffered no harm, denied their motions for preliminary and permanent injunctive relief, and, granting their motion to consolidate the preliminary injunction hearing with the trial on the merits, dismissed their suit. Ordinarily, when a district court consolidates a hearing on a motion for preliminary injunction with a trial on the merits, its findings of fact are reviewed for clear error and its legal conclusions de novo. Pinette v. Capitol Square Review & Advisory Bd., 30 F.3d 675, 677 (6th Cir.1994), aff'd, 515 U.S. 753, 115 S.Ct. 2440, 132 L.Ed.2d 650 (1995). Because the facts here are undisputed, we review the district court’s decision de novo. Moreover, “[w]hen the district court upholds a restriction on speech as constitutional, we conduct a de novo review of the facts.” Tucker v. California Dep’t of Educ., 97 F.3d 1204, 1209 n. 2 (9th Cir.1996); see also Daily Herald Co. v. Munro, 838 F.2d 380, 383 (9th Cir.1988) (explaining that the rule that “[w]e review de novo the district court’s application of law to facts on free speech issues ... reflects a special solicitude for claims that the protections afforded by the First [1072]*1072Amendment have been abridged”) (internal citations omitted).
Discussion
I. The Free Speech Issues
The Gentalas’ basic submission in this case is that the City’s categorical exclusion of events “in direct support of religious organizations” from Fund support impermissibly infringes on the free speech rights of religious organizations such as the Prayer Committee, based on their religious point of view.9 The parties have therefore, quite understandably, argued vigorously about the free speech aspects, drawing upon different strains of First Amendment doctrine concerning the role of government in supporting communicative activity.
The Gentalas, for example, maintain that the Civic Events Fund is, under the tripartite forum analysis we use to determine the government’s obligations when it permits or encourages the use of its property for communicative activity by private speakers, either a limited public forum or a nonpublic forum.10 In either case, the Gentalas seek to convince us, discriminating against groups in the distribution of Fund support on the basis that their speech expresses a religious viewpoint is unconstitutional. In support of this contention, the Gentalas cite, inter alia, Rosenberger, Widmar v. Vincent, 454 U.S. 263, 271, 102 S.Ct. 269, 70 L.Ed.2d 440 (1981), and Lamb’s Chapel v. Center Moriches Sch. Dist., 508 U.S. 384, 394, 113 S.Ct. 2141, 124 L.Ed.2d 352 (1993), cases which hold that the government, in administering a limited public forum, may not exclude some or all types of religious expression.11
Tucson, not surprisingly, is of the view that the cases the Gentalas rely upon are not pertinent at all to the speech issues in this ease. Rather, Tucson argues, the situation here fits into an entirely different mode of First Amendment speech analysis, applicable when the government is in some measure engaged in communicative activity, as it was in National Endowment for the Arts v. Finley, 524 U.S. 569, 118 S.Ct. 2168, 141 L.Ed.2d 500 (1998) (government as patron), Arkansas Educ. Television Comm’n v. Forbes, 523 U.S. 666, 118 S.Ct. 1633, 140 L.Ed.2d 875 (1998) (government as editor), and Rust v. Sullivan, 500 U.S. [1073]*1073173, 193, 111 S.Ct. 1759, 114 L.Ed.2d 233 (1991) (government as policy proponent). In all of those cases, some degree of viewpoint or qualitative selection criteria were permitted because the government was engaged not simply in providing a forum for private speech but in forwarding its own program through the speech of others. See, e.g., Forbes, 523 U.S. at 673-74, 118 S.Ct. 1633 (noting that the government sometimes assumes roles, such as public broadcaster or curriculum developer, in which its exercise of “editorial discretion” in choosing among prospective speakers is a speech activity on behalf of the government itself, not subject to forum analysis); Downs v. Los Angeles Unified Sch. Dist., 228 F.3d 1003, 1016-17 (9th Cir.2000) (holding that a school district could control the content of messages posted on high school bulletin boards “without the constraint of viewpoint neutrality”).
Each of these opposing characterizations of the City’s role in developing and expending its Civic Events Fund has some considerations in its favor. On the one hand, the City in this case, like the university in Rosenberger, is administering a fund providing in-kind services for a wide range of speakers. A large majority of speakers who meet the applicable Civic Events criteria were in fact funded during the period for which there is evidence in the record, so the very strong element of qualitative governmental selectivity involved in cases such1 as Finley (concerning the administration of the National Endowment for the Arts) or Forbes (concerning the editorial discretion necessary in making journalistic decisions) is not present here.
On the other hand, the criteria for funding here are more selective than in Rosen-berger, albeit not as selective and discretionary as in Finley and Forbes. That most projects are funded may evidence only self-selectivity by event organizers aware of the applicable criteria or, more probably, the availability of sufficient funds for the period in question, a circumstance that may not always prevail. Additionally, the- City’s policy goals here are not, one might conclude, similar to those of the university in Rosenberger; while the university was concerned with assuring a vigorous interchange of ideas among students as part of the educational process, the City is concerned with providing for its citizens and tourists events of certain kinds that the City believes enhance Tucson’s ambiance as an attractive place to live and visit. Because that is the goal, the City affirmatively identifies itself as the sponsor of funded events, placing its imprimatur on the events in a manner somewhat like the editor of an anthology does, while the university in Rosenberger did quite the opposite. See 515 U.S. at 841, 115 S.Ct. 2510.
Were we required to select among the mutually exclusive Free Speech Clause paradigms the parties propound-each leading under currently prevailing First Amendment doctrine to a different result concerning the City’s authority to exclude speakers from a government-provided subsidy because they are engaged in a prayer service-the choice would be a difficult one. It is a choice we need not make, however, in order to resolve this case. For, even assuming that the Gentalas have the better of the battle of the paradigms, that conclusion would not end the inquiry. Rather, as the Supreme Court and this court have indicated repeatedly, the government may decline to subsidize private religious speech when doing so would violate the Establishment Clause, or, put another way, avoiding a violation of the Establishment Clause is a sufficiently compelling reason to justify government’s exclusion of certain private speech in a forum otherwise dedicated to communicative activity.12 See, e.g., Rosen[1074]*1074berger, 515 U.S. at 837, 115 S.Ct. 2510; Capitol Square Revieio & Advisory Bd. v. Pinette, 515 U.S. 753, 761-62, 115 S.Ct. 2440, 132 L.Ed.2d 650 (1995);13 Lamb’s Chapel, 508 U.S. at 394, 113 S.Ct. 2141; Widmar, 454 U.S. at 271, 102 S.Ct. 269; Cole v. Oroville Union High Sch. Dist., 228 F.3d 1092, 1101 (9th Cir.2000), pet. for cert. filed Jan. 2, 2001; cf. Tucker, 97 F.3d at 1212. We therefore agree with the district court that if the City’s Establishment Clause justification for the Civic Events Fund’s religious exclusion is valid, then the exclusion is valid as well.14 It is to the Establishment Clause inquiry, therefore, that we now direct our attention.
II. The Establishment Clause
To embark on that inquiry, however, is not necessarily to simplify matters greatly, for modern Establishment Clause jurisprudence is famously indistinct. As a recent summary of the last century’s Religion Clause cases put it:
The Supreme Court is presently split closely on almost all important issues of ... separation of church and state.... Resolution of difficult issues in this area “depends on the hard task of judging-sifting through the details.... Such judgment requires courts to draw lines, sometimes quite fine, based on the particular facts of each case.” [Rosenberger, 515 U.S. at 847, 115 S.Ct. 2510 (O’Connor, J., concurring) ]. As a consequence of these realities, the current development of the law provides more themes than categorical principles, with little doctrinal stability.
Jesse H. Choper, A Century of Religious Freedom, 88 Cal. L.Rev. 1709, 1741 (2000). The most propitious approach to the perplexities of Establishment Clause doctrine, we have concluded, is to begin by examining three of the Establishment Clause “themes” involved both in Rosenberger, the case on which the Gentalas most heavily rely and which is indeed quite similar to this one in many respects, and in this case as well. We then turn to the details of Tucson’s Civic Events program to see why, as we hold, the Establishment Clause considerations play out differently on the facts of this case than they did in Rosenberger.
A. Establishment Clause Analysis in Rosenberger and Thereafter
In Rosenberger, the Supreme Court examined a Student Activities Fund (“Activities Fund”) program at the University of Virginia to determine whether Wide Awake Productions, an evangelical student-written newspaper otherwise eligible for an Activities Fund subsidy to cover its printing costs, could be denied the subsidy because the publication constituted a “religious activity.” The Activities Fund was financed through mandatory student activity fees; qualified student groups were required to submit their bills to the Student Council for payment from the Activities Fund. 515 U.S. at 824, 115 S.Ct. 2510. The Court concluded that the Activities Fund was, for purposes of Free Speech Clause analysis, a designated public forum, albeit “a forum more in a metaphysical than in a spatial or geographic sense,” from which the University could not ex-[1075]*1075elude speakers based upon their viewpoint. Id. at 830-31, 115 S.Ct. 2510. After determining that the exclusion of Wide Awake Productions was based on its Christian viewpoint rather than on its content, the Court held that, absent a compelling justification, denying Activities Fund support to Wide Awake would be a denial of the First Amendment’s Free Speech guarantee. Id. at 837,115 S.Ct. 2510. The Court then turned, as we have turned here, to the question whether the exclusion “is excused by the necessity of complying with the Constitution’s prohibition against state establishment of religion.” Id.
In doing so, the Court discussed three considerations that have been of great importance in recent Establishment Clause analysis, each of which has been further developed in subsequent cases:
First, the.majority opinion in Rosenber-ger emphasized that the Student Activities Fund program, if revised to include religiously-oriented publications, would not single out religion for special benefits, but, instead, would be “neutral toward religion.” 515 U.S. at 840, 115 S.Ct. 2510. At the same time, the Court in Rosenberger did not rest on the neutrality consideration. Instead, the Court observed that Establishment Clause analysis, as in comparable cases, turned not only on “general principles,” id. at 839, 115 S.Ct. 2510, namely the Activities Fund’s neutrality toward religion, but also on other “practical details of the program’s operation.” Id.
Moreover, neutrality is merely “one hallmark of the Establishment Clause,” as Justice O’Connor convincingly explained in Rosenberger, 515 U.S. at 846, 115 S.Ct. 2510 (O’Connor, J., concurring):
Our cases have permitted some government funding of secular functions performed by sectarian organizations, [citations omitted]. These decisions, however, provide no precedent for the use of public funds to finance religious activities..
This case lies at the intersection of the principle of government neutrality and the prohibition on state funding of religious activities.... Not to finance Wide Awake, according to petitioners, violates the principle of neutrality by sending a message of hostility toward religion. To finance Wide Awake, argues the University, violates the prohibition on direct state funding of religious activities.
When two bedrock principles so conflict, understandably neither can provide the definitive answer. Reliance on categorical platitudes is unavailing.
Id. at 847, 115 S.Ct. 2510.
Since Rosenberger, the Supreme Court has continued to struggle mightily over the question whether neutrality toward religion is sufficient, or nearly sufficient, to justify government subsidies of core religious activity, whatever else may be the case. A majority of the Court, however, made quite clear last term, if it was not clear before, that, although neutrality is an important consideration in Establishment Clause cases, that consideration alone is not determinative where government subsidy of religious activity is concerned.
Thus, Justice Souter, writing for three justices, after first reviewing at length the history of the neutrality concept in Establishment Clause jurisprudence, Mitchell v. Helms, 530 U.S. 793, 120 S.Ct. 2530, 2578-81, 147 L.Ed.2d 660 (2000) (Souter, J., dissenting), stated:
“Neutrality” [meaning] generality or ev-enhandedness of distribution ... is relevant ... but this neutrality is not alone sufficient to qualify the [government] aid as constitutional. It has to be considered only along with other characteristics of aid, its administration, its recipients, or its potential that have been emphasized over the years as indicators of just how religious the intent and effect of a given aid scheme really is.
Id. at 2581. Similarly, in her opinion for two justices in Mitchell, id. at 2557 (O’Connor, J., concurring in the judgment), Justice O’Connor decisively reiterated the position she took in her Rosenberger concurrence against neutrality as the deter[1076]*1076mining factor in Establishment Clause cases, stating: “[Neutrality is important but it is by no means the only ‘axiom in the history and precedent of the Establishment Clause.’ Rosenberger, [515 U.S. at 846, 115 S.Ct. at 2525 (O’Connor, J., concurring).] ... [Neutrality is not alone sufficient to qualify the aid as constitutional.”
We are left, then, with the clear holding by a Supreme Court majority that when the government subsidizes religious activity, the fact that it is doing so pursuant to a program that treats religious speech or association coequally with other speech or association is not, standing alone, determinative in Establishment Clause analysis. Instead, we are obligated to consider other factors in determining whether the connection between the government subsidy and the religious activity is such as to violate the concerns underlying the Establishment Clause (including, most importantly, the two we next discuss).15
Second, the central concern making Ro-senberger a close and difficult case was that the subsidy to religious expression involved was not simply the use of government space, generally open to the public or to some subgroup thereof for expressive activity, see Lamb’s Chapel, 508 U.S. at 395, 113 S.Ct. 2141; Widmar, 454 U.S. at 271-72 & n. 12, 102 S.Ct. 269, but a form of governmental financial support for religious activity. The Rosenberger Court, far from discounting this consideration, recognized that “[t]he Court of Appeals (and the dissent) are correct to extract from our decisions the principle that we have recognized special Establishment Clause dangers where the government makes direct money payments to sectarian institutions.” 515 U.S. at 842, 115 S.Ct. 2510. And indeed, the prohibition on funding religious activities has been preeminent among establishment of religion concerns, depending upon one’s view of history, from the nation’s infancy, see id. at 868-76, 115 S.Ct. 2510 (Souter, J., dissenting); see also id. at 846-47, 115 S.Ct. 2510 (O’Connor, J., concurring) (agreeing that the “axiom” that “ ‘[p]ublic funds may not be used to endorse the religious messages’ ” [citation omitted] has long been a “bedrock” of Establishment Clause jurisprudence); compare id. at 853-63, 115 S.Ct. 2510 (Thomas, J., concurring), or, at least, since Everson v. Board of Education, 330 U.S. 1, 17-18, 67 S.Ct. 504, 91 L.Ed. 711 (1941) (“No tax in any amount, large or small, can be levied to support any religious activities or institutions, whatever they may be called, or whatever form they may adopt to teach or practice religion.”).16
Rosenberger explicitly so recognized, noting that “if the State pays a church’s bills it is subsidizing it, and we must guard against this abuse.” 515 U.S. at 844, 115 S.Ct. 2510. The Rosenberger Court nonetheless held, for several reasons, that the [1077]*1077funding prohibition was not of controlling importance in that case, emphasizing that: (1) the mandatory fees comprising the Activities Fund were an “exaction upon the students,” not “a general tax designed to raise revenue for the University,”17 id. at 840-41, 115 S.Ct. 2510; see also id. at 841, 115 S.Ct. 2510 (“ ‘A tax, in the general understanding of the term, and as used in the Constitution, signifies an exaction for the support of the Government.’ ” (quoting United States v. Butler, 297 U.S. 1, 61, 56 S.Ct. 312, 80 L.Ed. 477 (1936))); id. (“Our decision, then, cannot be read as addressing an expenditure from a general tax fund.”); id. at 851-52, 115 S.Ct. 2510 (O’Connor, J., concurring); (2) Wide Awake was a “student publication[,] ... not a religious institution, at least in the usual sense of that term as used in our case law,” so that, in paying Wide Awake’s bills, the University was not paying the bills of a church or similar organization, id. at 844, 115 S.Ct. 2510;18 (3) the activity funded did not involve “religious exercises,” id. at 842, 115 S.Ct. 2510; and (4) the benefit provided did not involve direct funding of religious speech, since the University did not provide eligible student groups with direct financial assistance but instead paid their printing bills. Id. at 842-44, 115 S.Ct. 2510.
Third, also prominent in Rosenberger was the concern that subsidizing Wide Awake’s religiously-oriented publication could be perceived as a governmental endorsement of the religious ideas propounded by its student editors. The Court recognized that the Establishment Clause forbids “ ‘government speech endorsing religion,’ ” 515 U.S. at 841, 115 S.Ct. 2510 (quoting Board of Educ. of Westside Cmty. Schs. (Dist. 66) v. Mergens, 496 U.S. 226, 250, 110 S.Ct. 2356, 110 L.Ed.2d 191 (1990) (plurality opinion) (emphasis in original)), but concluded that, “[i]n this case, ‘the government has not fostered or encouraged’ any mistaken impression that the student newspapers speak for the University.” 515 U.S. at 841, 115 S.Ct. 2510 (quoting Capitol Square, 515 U.S. at 766, 115 S.Ct. 2440). In particular, the Court noted that “[t]he University has taken pains to disassociate itself from the private speech involved in this case,” Rosenberger, 515 U.S. at 841, 115 S.Ct. 2510; for example, the University required all recognized student organizations to sign an agreement stating that the recognition does not signal University “approval of the organizations’ goals or activities,” and also required each publication to include the statement that it was independent of the University, and that the University was not responsible for its actions. Id. at 824, 115 S.Ct. 2510; id. at 849, 115 S.Ct. 2510 (O’Connor, J., concurring). Given this agreement and disclaimer, as well as the fact that Wide Awake competed with many other Activities Fund-supported publications for the University community’s attention, it was extremely unlikely that observers would perceive that the University was endorsing the views put forward in Wide Awake.
Since Rosenberger, the precise nature of the Establishment Clause endorsement inquiry, like the role of the neutrality principle, has been clarified. Santa Fe centered on the endorsement question, stressing that the concern underlying the endorsé[1078]*1078ment inquiry is that governmental identification with religious speech or worship communicates to nonadherents “that they are outsiders, not full members of the political community, and an accompanying message to adherents that they are insiders, favored members of the political community.” 120 S.Ct. at 2279.19 To determine whether a government subsidy to religion constitutes “actual or perceived endorsement,” the Court stressed, requires a highly fact-specific inquiry, directed at the question “whether an objective observer, acquainted with the text, ... history, and implementation of the [program] would perceive it as a state endorsement” of religious observance. Id. at 2278.
Rosenberger and the Court’s post-Eo-senberger cases, then, provide a modicum of theoretical insight into the applicable Establishment Clause “themes,” and, more usefully perhaps, substantial practical insight into the specific factors most likely to be informative as we focus our attention on the interaction among these various considerations in the circumstances of this case.
B. “Sifting Through the Details”
As the foregoing analysis suggests, cases of this kind-that is, those at the intersection of the Free Speech and Establishment Clauses-are ones in which the admonition that Establishment Clause “jurisprudence ... is of necessity one of line-drawing,” Lee v. Weisman, 505 U.S. 577, 598, 112 S.Ct. 2649, 120 L.Ed.2d 467 (1992), has particular force. Turning to the line-drawing task, we conclude that this case involves significant factual differences from Rosenberger in matters that were deemed important to that case’s Establishment Clause analysis. There are also some contextual matters that we deem to be important but that had no counterpart in Rosenberger. The upshot is that, looking at these factors as a whole, we conclude that in this case, the public funding and government endorsement considerations considerably overwhelm the neutrality consideration, and lead us to the conclusion that Tucson’s Establishment Clause justification for its religious organization exclusion is valid.
(1) Tax-based funding of religious organization
Considering the governmental funding of religion, there are a number of factors that are precisely opposite in this case from a factor deemed important in Rosen-berger. Here, the source of the Civic Events Fund moneys includes general tax revenue, not simply fees levied against a discrete group of citizens. Also, Tucson excludes funding only for events supporting religious organizations, and the Prayer Committee is preeminently a religious organization; Wide Awake, in contrast, was a student newspaper with a religious viewpoint, not an “organization whose purpose is to practice a devotion to an acknowledged ultimate reality or deity.” Rosenberger, 515 U.S. at 840, 115 S.Ct. 2510 (quoting the university’s definition of “religious organization”) (emphasis added). Further, the activity funded here was a participatory religious service, led by ministers and consisting primarily of prayer, not a written document, however proselytizing. Unless we are to believe that the Rosenberger Court’s explicit reliance on each of these three considerations was without significance, we must conclude that in this case, unlike in Rosenberger, Civic Events Fund subsidies of the Prayer Day and similar events would be tantamount to “the State pay[ing] a church’s bills.” Id. at 844.
Moreover, Tucson’s administration of its Civic Events Fund involves considerably [1079]*1079more discretionary decisionmaking than did the university’s administration of the Activities Fund in Rosenberger. Here, the City has retained a good deal of subjective discretion concerning which events to fund, including establishing fairly general substantive criteria, criteria relating to the organization’s (and the Fund’s) financial circumstances, criteria related to whether the event is a first-time or ongoing event, and criteria concerning the likely appeal of the event to the citizenry. That most nonprofit organizations’ events are nonetheless funded does not detract from the fact that City officials do in fact review each application and make a decision using City-established criteria, that some applications are denied, and that the very existence of multiple criteria for funding likely leads to self-selection by otherwise eligible organizations. As a result, one cannot escape the conclusion, in this case, that it is the City making the choice which events to fund, a circumstance that, the Supreme Court has suggested, greatly exacerbates the significance of the tax-based derivation of the funding. See Mitchell, 120 S.Ct. at 2541-43 (plurality opinion) (stressing the importance of private, as opposed to governmental, funding choices as to whether tax-derived funds will be used by a religious institution pursuant to a program neutrally available to secular as well as religious institutions); id. at 2558-59 (O’Connor, J., concurring) (agreeing with the plurality’s stress on the importance of private choice, but maintaining that the private choice must be a true, individual choice, not a govemmentally-prescribed aid formula).
Finally, we note that Tucson’s 'exclusion for events in direct support of religious organizations appears to apply principally where, as here, the event will raise money from attendees that could be used to fund the organization in the future. The Genta-las, for example, explained their need for City funding in part by their desire to generate a surplus from the “free will offering.” There is therefore the additional concern, not present in Rosenberger or in any other case concerning governmental funding of religious activity of which we are aware, that although the subsidy provided is in-kind and in that sense indirect, it is intended for use effectively as seed money to generate other, non-earmarked funds for the religious organization. The exclusion for events “in direct support of religious organizations” therefore functions similarly to a prohibition on cash grants to religious organizations. Rosenberger, 515 U.S. at 842, 115 S.Ct. 2510.
(2) City Endorsement of Religion
Turning to the question “whether an objective observer, acquainted with the text, ... history, and implementation of the [program] would perceive it as a state endorsement” of any religious observance funded, Santa Fe, 120 S.Ct. at 2279, we conclude that there are considerably more reasons that such an observer would so conclude regarding Tucson’s Civic Events Fund than there were in Rosenberger.
Here, the subsidy often includes, as it would have for the Prayer Day event, the on-site, observable presence of city employees, not simply the behind-the-scenes payment of bills from private vendors, as in Rosenberger. Moreover, the equipment used — microphones, lighting systems, tables, public address systems — is city-owned and is likely to be quite visible to onlookers during the event.
Additionally, the detailed, subjective funding criteria discussed above, as well as the role of the Mayor and Council in establishing and applying those criteria, would contribute to an informed observer’s impression that an event that survives the review process is one that the City affirmatively endorses. Contributing to that impression would be the fact that many, probably most, events in the parks do have to pay the prescribed fees because they are not eligible for Civic Events' Fund designation. In short, the City has retained a role for itself something like that of impresario, selecting the events that are consistent with the image of Tucson that [1080]*1080the Cr y wishes to foster and that therefore merit public subsidy.
Consistent with the impression created by the funding process, the City’s policy and application form make clear that the Civic Events Fund program is meant to endorse some events as “Civic Events” worthy of the City’s imprimatur. A person acquainted with the funding program’s text, history, and implementation would know, for example, that the governing policy statement expressly states that “[t]he Mayor and Council encourage and support Civic Events,” (emphasis added), which are then described by category; that the phrase “Civic Events” is always capitalized in the policy statement even when no funding is at issue, indicating a city-conferred title or status; and that organizers of funded events are directed to deal with a city official denominated the “Civic Event Coordinator” in order to “obtain assistance in coordinating the event.” Of a piece with these other indicia of endorsement is the additional, significant requirement that, far from declaring their independence from the government funder as in Rosenberger, organizers of Civic Events Fund-subsidized events must call attention to the City’s supportive role both in pre-event advertising and at the event itself.
It is true, as the district court noted in discounting the significance of the latter point, that the City could simply remove this acknowledgment requirement if it chose. That observation is not pertinent to our analysis, for two related reasons.
First, as the other factors summarized above indicate, the express acknowledgment requirement simply confirms the overall nature of the Civic Events program as one in which the City affirmatively supports certain privately-originated events. The acknowledgment requirement does not itself create that relationship. To require the City to hide its true relationship to the event would not alter that relationship, but instead, would only mislead the public.
Second, the courts are not in the business of designing governmental programs but, instead, of taking them as they find them and determining their constitutional significance. Thus, for example, the Court in Forbes, 523 U.S. 666, 118 S.Ct. 1633, 140 L.Ed.2d 875, did not direct the government henceforth to run its broadcasting stations on an open microphone basis so as to avoid editorial decisions but, instead, viewed the broadcasting program as one in which editorial decisions were important and came to its conclusions about the Free Speech rights of those seeking access to the broadcasting station accordingly. Nor did the Court solve the knotty Free Speech issues regarding government funding for the arts involved in Finley, 524 U.S. 569, 118 S.Ct. 2168, 141 L.Ed.2d 500, by informing the government that it should fund the arts on a first-come-first-served basis. See id. at 598, 118 S.Ct. 2168 (Scalia, J., concurring in the judgment) (“It is the very business of government to favor and disfavor points of view on ... innumerable subjects.... [Government] furtherfs] [a] favored point of view by achieving it directly, ... by advocating it officially, ... or by giving money to others who achieve and advocate it.”); Downs, 228 F.3d at 1013-14. To require Tucson to drop its express support of preferred community events in favor of an objective, hands-off funding system in order that events in direct support of religious organizations could be funded would be to require the City to abandon its present program and substitute a different one, a ukase beyond our authority as a court.
We note, as well, that the nature of Tucson’s Civic Events Fund program is such that allowing various religious organizations to compete for funding would lead to some of the very dangers the Establishment Clause was intended to prevent. For one thing, in order to comply with the City’s emphatically secular criteria for funding, religious event organizers would be likely to “secularize” their events. Lee, 505 U.S. at 608, 112 S.Ct. 2649 (Blackmun, J., concurring). Although the focus on the [1081]*1081community’s “historical, cultural, and ethnic heritage,” “critical issues,” and “civic pride” may be commendable, creating an incentive to structure religious events toward such worldly objectives would encourage the dilution of spiritual life, not enhance it. And while fairly small sums are at issue here, the Establishment Clause principles examined above admit no distinction between large grants to religion and small ones.
Further, the necessity of applying subjective substantive criteria could result in the City impermissibly promoting some faiths over others. Lee, 505 U.S. at 627, 112 S.Ct. 2649 (Souter, J., concurring); Everson, 330 U.S. at 15, 67 S.Ct. 504. Some religions are likely to have more “broad community appeal” than others, or to “contribute to tourism” more readily. Concerns about the dangers of such government favoritism to the cohesiveness of civil society echo down from the Framers to the Supreme Court’s most recent jurisprudence. Compare James Madison, Memorial and Remonstrance Against Religious Assessments, ¶ 9 (1785) (observing that government favoritism toward certain faiths “degrades from the equal rank of Citizens all those whose opinions in Religion do not bend to those of the Legislative authority”), with Santa Fe, 120 S.Ct. at 2279 (noting that government sponsorship of a religious message communicates to nonadherents “ ‘that they are outsiders, not full members of the political community’ ” (quoting Lynch v. Donnelly, 465 U.S. 668, 688, 104 S.Ct. 1355, 79 L.Ed.2d 604 (1984) (O’Connor, J., concurring))).
The short of the matter is that, after “sifting through the details,” we are convinced, first, that this case and Rosenber-ger, for all their superficial similarity, are different in most of the details that matter for Establishment Clause purposes, and, second, that all of the ways in which they differ point toward an Establishment Clause violation in this case.20
C. Relevance of Free Exercise Clause to the Establishment Clause
One final point: We do not agree that, as Judge Fernandez’ dissenting opinion maintains, Free Exercise Clause considerations require that we alter our Establishment Clause analysis.
Denying financial subsidies to religious organizations does not, standing alone, substantially burden Free Exercise rights. As the Supreme Court observed in Bob Jones University v. United States, 461 U.S. 574, 603-04, 103 S.Ct. 2017, 76 L.Ed.2d 157 (1983), religious schools’ ineligibility for a tax exemption due to their racially discriminatory practices might impose significant burdens on their operations, but any such administrative burdens would “not prevent those schools from observing their religious tenets.” See also Hernandez v. Commissioner, 490 U.S. 680, 699, 109 S.Ct. 2136, 104 L.Ed.2d 766 (1989) (denial of tax deduction not a Free Exercise Clause violation); Swaggart Ministries v. California Dep’t of Equalization, 493 U.S. 378, 391-92, 110 S.Ct. 688, 107 L.Ed.2d 796 (1990) (same).
[1082]*1082Here, the claim of denial of Free Exercise rights is particularly weak. The Gentalas have made no showing that their ability to worship as they choose has been discernibly burdened by the operation of the Civic Events program. Rather, the Prayer Committee staged a well-attended prayer service in a City-provided location, the Reid Park bandshell. See Collins v. Chandler Unified Sch. Dist., 644 F.2d 759, 763 (9th Cir.1981). In Free Exercise terms, then, the government’s denial of a financial subsidy for microphones and lights to a religious group is a far cry from its outright prohibition of the group’s religious practice. Cf. Church of the Lukumi Babalu Aye, Inc. v. City of Hialeah, 508 U.S. 520, 113 S.Ct. 2217, 124 L.Ed.2d 472 (1993) (holding that ordinance prohibiting a religious group’s ritual violated the Free Exercise Clause).
Moreover, the refusal to fund the Prayer Committee’s microphones and hghting is also a far cry from denying a religious organization or event services essential to survival in civil society, such as police and fire protection. Police and fire protections are vital pubhc services to which all citizens enjoy a basic entitlement. Denying such services to a particular group of citizens would both evince government hostility and threaten the group’s abihty to exist. The same simply cannot be said of Tucson’s program, a subsidy for items such as microphones and stage lights for special events, available only to a hmited set of potential users and in no way necessary for survival in the modern world. Indeed, if the failure to subsidize a religious activity with regard to expenditures that are in no way analogous to basic pubhc safety services violated the Free Exercise Clause, the Court would have said so in Rosenber-ger and written a very different, and much shorter, opinion.
We therefore conclude that Tucson was correct that it could not provide funding to “events in direct support of religious organizations” generally, or to the Prayer Day event in particular, without violating its constitutional obligations regarding the separation of church and state.
Conclusion
Because we agree with the City that its' policy is mandated by the Estabhshment Clause, we need not reach the question that the City raises on cross-appeal: whether the district court should have granted the City leave to amend its answer to add an affirmative defense derived from the Arizona Constitution’s religion clauses.21 The federal Constitution provides all the defense Tucson needed in this lawsuit. We affirm the district court in all respects.
AFFIRMED.