ON REHEARING EN BANC
FERREN, Associate Judge:
In mid-November 1987, appellants Fedo-rov 1 and Donne were arrested for refusing to leave the Farragut West Metro station after hours. They were engaged in a political demonstration protesting erection of a fence and locked gate to keep out the homeless at night. Appellant Mellecker was arrested in early December 1987 for refusing to leave the Farragut West station during a similar demonstration. Each appellant was charged with one count of unlawful entry for failing to quit the Metro station on demand of lawful authority. D.C.Code § 22-3102 (1989).2 After their arrests, all three appellants — students at George Washington University — were denied admission to the United States Attorney’s pretrial diversion program although each, as a first-time offender, was eligible for diversion.3 Fedorov and Donne argued unsuccessfully before Judge Salzman, and Mellecker argued fruitlessly to Judge Burnett, that the United States Attorney had a policy of denying diversion to political demonstrators and that this policy constituted both selective prosecution, in violation of the Fifth Amendment, and an impermissible penalty on speech, in violation of the First Amendment. This appeal presents the question whether appellants have proffered sufficient evidence to entitle them to discovery and to an evidentiary hearing on their constitutional claims. We conclude that appellants’ proffers reflect a prima facie case of both alleged constitutional violations. Accordingly, we reverse and remand for further proceedings.
I.
A. Veronica Fedorov and Stephanie Donne
On the evening of November 18, 1987, appellants Fedorov and Donne arrived at the Farragut West Metro station to partic[373]*373ipate in a political demonstration against the Washington Metropolitan Area Transit Authority. The demonstrators opposed a relatively new Metro policy of barring homeless people from the station area after hours by enclosing the Metro station with a locked gate. At closing time, Metro police warned the demonstrators that they must move from the gate or suffer arrest. Fe-dorov and Donne remained in the area and were arrested, arraigned, and charged with unlawful entry under D.C.Code § 22-3102, supra note 2. Although defense counsel applied for pretrial diversion, see supra note 3, the requests were denied.
On December 22, 1987, Fedorov and Donne filed a motion to dismiss. First, they claimed selective prosecution, relying on the two-part test recognized in Wayte v. United States, 470 U.S. 598, 605-06, 105 S.Ct. 1524, 1529-30, 84 L.Ed.2d 547 (1985), and in (Elizabeth) Smith v. United States, 460 A.2d 576, 578 (D.C.1983). They alleged that (1) others “similarly situated” — i.e., other eligible first-time offenders charged with unlawful entry — had been diverted from the criminal justice system, while they and all other charged political demonstrators had been prosecuted, and that (2) the government’s refusal to divert political demonstrators was based on invidious discrimination against the exercise of First Amendment rights, resulting in a denial of equal protection under the Fifth Amendment. Fedorov and Donne also claimed a direct violation of their First Amendment rights, arguing that the government withheld a valuable benefit — pretrial diversion — because of their political speech. They requested an evidentiary hearing and, later, filed motions to compel discovery.
In their motions, Fedorov and Donne proffered that (1) they were first offenders who had no other pending charges or prior arrests and thus were eligible for diversion; (2) they were “similarly situated” with all persons charged with unlawful entry who were otherwise eligible for diversion; (3) preliminary results of an investigation into the approximately two thousand cases of unlawful entry over a three-year period (1985-87) had revealed that eight non-political violators had been granted diversion but that no political demonstrators had received that treatment; (4) student counsel for appellants had contacted Katherine Ellis, the Assistant United States Attorney handling the cases, and Ellis had informed counsel that the government had decided “to deny diversion to anyone arrested during the demonstrations at the Farragut West Metro Station;" (5) Debbie Jones, a staff member of the United States Attorney’s Office, informed Fedorov without explanation that she need not attend her pretrial diversion conference because she would be denied diversion; and (6) Katherine Winfree, Chief of the Misdemeanor Trial Section in the United States Attorney’s Office, had informed student counsel in another Farragut West Metro case that the government had denied diversion to a Farragut West defendant because “he had engaged in a political demonstration.”
The government replied that Fedorov and Donne had failed to carry their heavy burden to establish a prima fade showing of selective prosecution because the appropriate group of “similarly situated” persons, for comparison purposes, was the group of all those arrested at Farragut West on the night of November 18, 1987, and all those demonstrators had been prosecuted. In short, the government contended that the prosecution of all the Farragut West demonstrators proved that the government had a policy of treating all similarly situated offenders alike.
As to appellants’ second argument, the government focused on the conduct underlying the prosecutions, arguing that the First Amendment was not implicated because Fedorov and Donne had violated a content-neutral statute and the First Amendment did not protect their right to remain unlawfully on government property. The government added that these appellants had not offered any evidence suggesting a reason why the government would want to silence protesters on behalf of the homeless, rather than remain neutral in its enforcement of the unlawful entry statute.
During a status hearing on January 28, 1988, Judge Salzman denied appellants’ mo[374]*374tions, refusing even their request to make an offer of proof, and noting that appellant’s position was “utterly without legal merit.” In his written memorandum and order of February 4,1988, the judge explicitly adopted the government’s position. He concluded that Fedorov’s and Donne’s definition of “similarly situated” — essentially all charged first-time offenders under the statute — was too broad because “an individual prosecuted for failing to leave a department store after previously being barred in lieu of prosecution for shoplifting is not ‘similarly situated’ to these defendants, though charged with the identical statutory offense of unlawful entry.” Memorandum and Order, Salzman, J. at 3 (Feb. 4, 1988). The judge accordingly determined that only Farragut West demonstrators on behalf of the homeless on the day in question were “similarly situated” with appellants. Because the government had denied all Farragut West demonstrators diversion, the judge then ruled that Fedorov and Donne had failed to show they had received disparate treatment. As a result, their selective prosecution claim failed.4
Judge Salzman also concluded that Fedo-rov and Donne had not made a colorable showing of a First Amendment violation because, at the time of their arrest, Metro closed the Farragut West station “for all purposes,” not just to curtail Fedorov’s and Donne’s expressive activities. (Emphasis in original).
B. Dana Mellecker
Appellant Mellecker was charged with one count of unlawful entry, D.C.Code § 22-3102, supra note 2, for failing to quit the Farragut West Metro station during a demonstration on behalf of the homeless on December 5, 1987. When he, too, was denied diversion, see supra note 3, he filed a motion to dismiss for selective prosecution, requesting an evidentiary hearing and raising the same challenges that Fedorov and Donne had raised. The motion relied on the Wayte and (Elizabeth) Smith test for selective prosecution and contained proffers similar to those submitted in Fedorov’s and Donne’s motions: that (1) Mellecker was eligible for diversion; (2) he was “similarly situated” with all those charged with unlawful entry and eligible for diversion; (3) a paralegal at the United States Attorney’s office had stated that that office had a policy against diverting the Farragut West demonstrators; (4) Mellecker’s diversion interview had been perfunctory; and (5) Ms. Winfree, Chief of the Misdemeanor Trial Section of the United States Attorney’s Office, had informed student counsel in another Farragut West case that the United States Attorney had “a policy against diverting individuals who engage in political demonstration.”
Mellecker also filed supplemental information which more fully described the study of unlawful entry cases that Fedorov and Donne had cited. According to this information, the D.C. Law Students in Court program had examined the court jacket in every unlawful entry case in the District of Columbia from January 1, 1985 to January 1, 1988. During that period, approximately 2,000 persons had been arrested for unlawful entry. Forty percent were “no-papered” or dismissed while others were papered on other charges. The remaining sample of 953 cases consisted of 275 persons classified as political demonstrators, 648 classified as non-demonstrators, and 30 who could not be classified. Of the 275 political demonstrators, none (0%) was admitted to diversion; of the 648 non-demonstrators, at least 27% were [375]*375granted diversion.5
At a March 29, 1988 hearing, Mellecker’s counsel proffered her own statement in support of the allegation that the United States Attorney had a policy of not diverting otherwise eligible unlawful entrants if they were political demonstrators:
The effort was made by me, as counsel for Mr. Mellecker ..., to get [this] Defendant, who, as I say, ha[s] no prior record, no other pending charges, and who [is] charged with unlawful entry, a normally divertable offense, into the diversion program. However, I was advised by an employee of the United States attorney’s office that it would be a waste of time, essentially, to bring Mr. Mellecker ... in for [a] diversion interview, because I was told over the telephone, the United States attorney’s office has a policy of not diverting protest cases[.]
She also proffered the testimony of two witnesses. The first, a student attorney representing another Farragut West demonstrator, was prepared to testify that Ms. Winfree, the Chief of the United States Attorney’s Misdemeanor Trial Section, had informed the student attorney that the government has “a policy of not diverting protest cases.” This witness would also testify about the methodology the D.C. Law Students in Court program had used in its statistical study of the government s diversion practices and about the reliability of this method. The second witness, a statistics professor from Howard University who had assisted in the development of the study and had analyzed the results, would testify about the statistical significance of the finding that not a single political demonstrator charged with unlawful entry had been granted diversion while at least 27% of charged non-demonstrators had been admitted to diversion.
At the hearing, the government opposed Mellecker’s request for an evidentiary hearing. In response to repeated inquiries from Judge Burnett, the prosecutor would not categorically confirm or deny that the government had a policy of denying diversion to political demonstrators.6 When the judge directly asked the government whether any political demonstrator in the last three years had been admitted to pretrial diversion, the government responded that the office does not maintain statistics and “we don’t think that it matters whether there has been [a diverted political demonstrator] or not.” The government maintained, rather, that Mellecker had failed to show he was a victim of disparate treatment because those with whom he was similarly situated — those “arrested the same day that [he was] arrested,” — had also been denied diversion.7
[376]*376On April 14, 1988, Judge Burnett issued a memorandum opinion denying the motion to dismiss. Explicitly adopting Judge Salzman’s analysis in Fedorov’s and Donne’s cases, Judge Burnett rejected Mel-lecker’s selective prosecution argument, concluding that his proposed class of those “similarly situated” — those charged with unlawful entry who were technically eligible for diversion — “was far too broad.” Rather, the judge agreed with the government that the appropriate comparison group was “those who attempted to bar the closing of the Farragut West Metrorail station.” Memorandum, Opinion and Order, Burnett, J. at 3 (Apr. 14, 1988). The judge also concluded that, even if the appropriate comparison group should have been broader, Mellecker had failed to show that the government’s decision to prosecute was based on invidious discrimination. Judge Burnett reasoned that the government had prosecuted Mellecker because he had refused to leave the Metro station after repeated warnings, not because of the nature or content of his protest activities.
The judge used the same reasoning to reject Mellecker’s First Amendment claim, concluding that Mellecker had “made no showing that the Government’s policy with regard to its administration of its pretrial diversion program is intended to keep the defendant[] from protesting on behalf of the homeless or that, it is in any way based on the content or nature of the message communicated in the demonstration.”
Appellants Fedorov, Donne, and Mellecker subsequently entered conditional guilty pleas on one count of unlawful entry, preserving their rights to appeal from the denials of their motions to dismiss. Judge Salzman sentenced Fedorov and Donne to six months of probation on condition that they complete fifty hours of community service. Judge Burnett placed Mellecker on one year of probation, to be unsupervised after he had completed twenty-five hours of community service.
II.
Appellants frankly acknowledge that they have violated a criminal statute. They seek reversal of their convictions not because they claim innocence but because they allege a defect of constitutional magnitude in the initiation of their prosecutions. See United States v. Berrigan, 482 F.2d 171, 174-75 (3d Cir.1973). They allege, first, that the United States Attorney’s policy of denying diversion to those engaged in political demonstrations discriminates against a subclass of offenders on the basis of an impermissible classification: the exercise of the First Amendment right of political protest.
A.
Appellants’ claim is very difficult to prove because the government has broad discretion whether to prosecute an accused. “[S]o long as the prosecutor has probable cause to believe that the accused committed an offense defined by statute, the decision whether or not to prosecute, and what charge to file or bring before a grand jury, generally rests entirely in his [or her] discretion.” Bordenkircher v. Hayes, 434 U.S. 357, 364, 98 S.Ct. 663, 668, 54 L.Ed.2d 604 (1978). Courts are properly hesitant to examine the government’s decision to prosecute because that decision “is particularly ill-suited to judicial review.” Wayte, 470 U.S. at 607, 105 S.Ct. at 1530. “Such factors as the strength of the case, the prosecution’s general deterrence value, the Government’s enforcement priorities, and the case’s relationship to the Government’s overall enforcement plan are not readily susceptible to the kind of analysis the courts are competent to undertake.” Id.
Nonetheless, although prosecutorial discretion is broad, “it is not ‘unfettered. Selectivity in the enforcement of criminal laws is ... subject to constitutional constraints.’ ” Id. at 608, 105 S.Ct. at 1531 (quoting United States v. Batchelder, 442 U.S. 114, 125, 99 S.Ct. 2198, 2204, 60 L.Ed.2d 755 (1979)). “[T]he decision to prosecute may not be ‘deliberately based upon an unjustifiable standard such as race, religion, or other arbitrary classification,’ Bordenkircher, [434 U.S.] at 364 [98 S.Ct. at 668], quoting Oyler v. Boles, 368 U.S. 448, 456 [82 S.Ct. 501, 505, 7 L.Ed.2d [377]*377446] (1962), including the exercise of protected statutory and constitutional rights, see United States v. Goodwin, [457 U.S. 368, 372, 102 S.Ct. 2485, 2488, 73 L.Ed.2d 74 (1982)].” Wayte, 470 U.S. at 608, 105 S.Ct. at 1531. The basic principle was stated long ago in Yick Wo v. Hopkins, 118 U.S. 356, 6 S.Ct. 1064, 30 L.Ed. 220 (1886):
Though the law itself be fair on its face and impartial in appearance, yet, if it is applied and administered by public authority with an evil eye and an unequal hand, so as practically to make unjust and illegal discriminations between persons in similar circumstances, material to their rights, the denial of equal justice is still within the prohibition of the Constitution.
Id. at 373-74, 6 S.Ct. at 1072-73. Claims of selective prosecution, therefore, should be judged according to equal protection standards. See Wayte, 470 U.S. at 608, 105 S.Ct. at 1531 (appellant must show prosecutor’s enforcement had discriminatory effect and was motivated by discriminatory purpose).
The government’s decision to deny an arrestee admission into a diversion program is a decision to prosecute and we treat it as a claim of selective prosecution. See, e.g., Chavez v. United States, 499 A.2d 813, 815 (D.C.1985); Irby v. United States, 464 A.2d 136, 141 (D.C.1983); (Elizabeth) Smith, 460 A.2d at 578. Thus, the prosecutor’s decision not to grant diversion is subject to judicial review applying equal protection principles. See (Elizabeth) Smith, 460 A.2d at 578. A party who charges discriminatory application of prose-cutorial discretion to pretrial diversion "carries a heavy burden of proof,” id., for “[t]he conscious exercise of some selectivity in enforcement is not in itself a federal constitutional violation.” Oyler, 368 U.S. at 456, 82 S.Ct. at 506. Rather, the defendant must make a prima facie showing that: “(1) others similarly situated were not prosecuted, and (2) the selective prosecution being complained of was improperly motivated, i.e., it was based on an impermissible consideration such as race or a on a desire to prevent the exercise of constitutional rights.” (Elizabeth) Smith, 460 A.2d at 578; see Wayte, 470 U.S. at 608, 610, 105 S.Ct. at 1531, 1532 (“discriminatory purpose”); United States v. (James) Smith, 354 A.2d 510, 513 (D.C.1976) (“invidious or otherwise impermissible form of discrimination”).8
To support a claim of selective prosecution, both allegations must be clearly and separately established. See Attorney General of the United States v. The Irish People, Inc., 221 U.S.App.D.C. 406, 410, 684 F.2d 928, 932 (1982), cert. denied, 459 U.S. 1172, 103 S.Ct. 817, 74 L.Ed.2d 1015 (1983). The test is a rigorous one, see United States v. Mangieri, 224 U.S.App.D.C. 295, 298, 694 F.2d 1270, 1273 (1982), for “[t]he presumption is always that a prosecution for violation of a criminal law is undertaken in good faith and in nondiscriminatory fashion for the purpose of fulfilling a duty to bring violators to justice.” United States v. Falk, 479 F.2d 616, 620 (7th Cir.1973) (en banc). Appellants would be entitled to relief, however, “if [their] evidence proved that the authorities purposefully discriminated against those who chose to exercise their First Amendment rights.” United States v. Steele, 461 F.2d 1148, 1151 (9th Cir.1972).
To obtain discovery and an evidentiary hearing on a selective prosecution claim, a defendant is not required to prove each element of the claim. The caselaw, however, is not clear on precisely what showing must be made. One line of cases requires the defendant to make a prima facie showing9 of selective prosecution in order to [378]*378obtain discovery or an evidentiary hearing,10 see (Elizabeth) Smith, 460 A.2d at 578,11 while another line of cases would grant a motion for discovery or an eviden-tiary hearing upon a “colorable showing” that the two-element test had been met.12 See, e.g., Irish People, 221 U.S.App.D.C. at 410 & n. 8, 684 F.2d at 932 & n. 8; see also Wayte, 470 U.S. at 621 n. 1, 105 S.Ct. at 1538 n. 1 (Marshall, J., dissenting) (discussing conflict in federal circuits over standards permitting discovery). Still other cases have required different threshold showings by the defendant — a “nonfrivo-lous showing”13 or facts “tending to show” selective prosecution and raising “reasonable doubt about the propriety of the prosecutor’s purpose”;14 but these cases allow the government to present an “explanation” or “countervailing reasons” for its [379]*379conduct before the court grants discovery or an evidentiary hearing. See supra notes 13 and 14.
In this case, we need not resolve the precise contours of the showing a defendant must make to obtain discovery or an evidentiary hearing. Assuming, as the government contends, that a prima facie showing as to both criteria — “similarly situated” and "discriminatory purpose” — is required, we are satisfied (as elaborated below) that appellants have made such a showing. We conclude, accordingly, that appellants will be entitled to reasonable discovery and, thereafter, to an evidentiary hearing, subject to one condition attributable to the peculiar circumstances of this case: before discovery begins, the government shall have an opportunity upon remand to try to provide an explanation in rebuttal that vitiates appellants’ prima fa-cie showing. We now explain.
B.
To meet the first element of a selective prosecution case, a defendant must show that other similarly situated persons “are generally not prosecuted for the same conduct.” United States v. Aguilar, 883 F.2d 662, 706 (9th Cir.1989) (quoting United States v. Wilson, 639 F.2d 500, 503 (9th Cir.1981)), cert. denied, — U.S. -, 111 S.Ct. 751, 112 L.Ed.2d 771 (1991)); see also United States v. Penagaricano-Soler, 911 F.2d 833, 837 (1st Cir.1990); St. German of Alaska Eastern Orthodox Catholic Church v. United States, 840 F.2d 1087, 1095 (2d Cir.1988). The purpose of identifying a similarly situated class of offenders is to test whether the government is engaging in a pattern of selectivity by comparing prosecutions of offenders who engage in the exercise of their constitutional rights with prosecutions of those who do not. Absent a suitable explanation, the prosecution of all persons charged with a particular crime while exercising their constitutional rights, compared with prosecution of only a small proportion of those persons who are not, gives rise to an inference of discrimination. See Penagaricano-Soler, 911 F.2d at 838 (“statistical disparity in relative incidence” of discriminatory and non-discriminatory prosecutions sufficient to raise prima facie case).15
In this case, appellants argue that the appropriate control group is all those charged with unlawful entry who are technically eligible for diversion. See supra note 3. They contend that because the legislature saw fit to create this particular, undifferentiated category of offenders, see supra note 2 — as narrowed by the executive’s creation of a subcategory eligible for diversion — all eligible persons so defined {i.e., all first-time unlawful entrants without'Other pending charges) are “similarly situated.” In other words, they say, the unlawful conduct defined by the elements of the crime itself comprises the defining criterion for selective prosecution analysis, narrowed only by the eligibility requirements for diversion itself.
Appellants bolster their argument by challenging the government to cite a single case where the term “similarly situated” has not included all persons charged with violating a particular statute. See, e.g., Wayte, 470 U.S. at 602-04 & n. 3, 610, 105 S.Ct. at 1527-28 & n. 3, 1532 (all reported to have violated Presidential Proclamation No. 4771 by failing to register for draft); United States v. Matter, 818 F.2d 653, 655 [380]*380(8th Cir.1987) (all charged with violating 18 U.S.C.App. § 1202(a)(1) for possessing a firearm after previous conviction of felony); Irish People, 221 U.S.App.D.C. at 409, 684 F.2d at 931 (all who failed to register under Foreign Agents Registration Act); United States v. Scott, 521 F.2d 1188, 1195 (9th Cir.1975) (all charged with violating 26 U.S.C. § 7203 for failure to file income tax returns), cert. denied, 424 U.S. 955, 96 S.Ct. 1431, 47 L.Ed.2d 361 (1976); Steele, 461 F.2d at 1150 (all charged with violating 13 U.S.C. § 221(a) by refusing to answer census); Smith, 460 A.2d at 578 (all 56 persons arrested for violating D.C.Code § 9-112(b)(4) during 1973-1981 for unlawful conduct on United States Capitol Grounds); United States v. Wilson, 342 A.2d 27, 30 n. 5 (D.C.1975) (all 550 persons arrested during 1973 and charged with violating D.C.Code § 22-2701 for solicitation for prostitution).
Both trial judges, however, concluded that the appropriate comparison group for selective prosecution analysis was all persons who had participated the same night in the demonstration on behalf of the homeless at the same place (Farragut Station). This classification is fatally underinclusive, however, for it begs the question. By so narrowing the class of similarly situated offenders, the trial judges defined the critical constitutional issue out of the case. Instead of creating a control group that could be used to determine whether a defect of constitutional magnitude could be found in the initiation of appellants’ prosecutions, Judge Salzman and Judge Burnett, in effect, eliminated all room for comparison by considering as the control group only those persons whose constitutionally significant behavior was the same as that of appellants. In other words, the trial judges focused their inquiry on whether there was disparate treatment within the class of alleged victims of discrimination, rather than on whether those alleged victims, taken as a group, were treated less favorably than others who were similarly situated except for the exercise of protected rights.
To use an imperfect but revealing analogy, under the analysis used by the trial judges a claim of racial disparity between a school in a predominantly white neighborhood and another in a predominantly African-American neighborhood would be resolved not by comparing the two but, rather, by confining the analysis to determining whether there was discrimination within the African-American school. Similarly, if the point of a defendant’s claim (and the trial court’s analysis) is to determine whether defendants exercising free speech rights are denied consideration for the diversion program, while other defendants not exercising those rights do receive diversion, then the trial court cannot properly compare only those who exercise First Amendment rights with others who exercise the same rights.
In sum, we conclude that the trial judges’ definition of “similarly situated” to include only those persons arrested at the same time and place as appellants is much too narrow and would inevitably defeat any meaningful comparison for selective prosecution purposes. Rather, we conclude that prima fade, i.e., presumptively, all persons charged with violating a particular statute (who are first-time offenders without other pending charges) are “similarly situated” for selective prosecution/diversion analysis. We are not prepared to hold that this must be the appropriate comparison group in all cases, for there may be considerations not presented here; but, we believe there should be a rebuttable presumption that everyone within the group of statutory offenders — here, every person charged with unlawful entry of any type— is “similarly situated” for selective prosecution purposes.16 Accordingly, the [381]*381government has the burden to show why this statutory class of offenders, which the legislature has intentionally grouped together, should not be the classification used to define “similarly situated” persons for selective prosecution analysis (as narrowed by reference to diversion eligibility). By limiting its argument in the trial court to comparing appellants only with other political demonstrators arrested the same night at Farragut West, the government has not met that burden. Prima facie, appellants are similarly situated with all other charged first-time offenders under the unlawful entry statute.
C.
The second element of a selective prosecution analysis requires a showing that the decision to prosecute the defendant was improperly motivated. “ ‘Discriminatory purpose’ ... implies more than ... intent as awareness of consequences. It implies that the decisionmaker ... selected or reaffirmed a particular course of action at least in part ‘because of,’ not merely ‘in spite of,’ its adverse effects upon an identifiable group.” Wayte, 470 U.S. at 610, 105 S.Ct. at 1532 (quoting Personnel Admin’r of Mass. v. Feeney, 442 U.S. 256, 279, 99 S.Ct. 2282, 2296, 60 L.Ed.2d 870 (1979)). A showing of discriminatory purpose, however, is not necessary when the selective prosecution claim “is based on an overtly discriminatory classification.” Wayte, 470 U.S. at 608 n. 10, 105 S.Ct. at 1531 n. 10 (citing Strauder v. West Virginia, 100 U.S. 303, 25 L.Ed. 664 (1880)). Stated conversely, discriminatory purpose need only be shown when a facially neutral policy is alleged to have discriminatory impact. An “overtly discriminatory classification,” therefore, is presumptively invalid. Unless the government shows that such apparent discrimination is not in fact taking place, the courts must engage in strict scrutiny to determine whether the government can prove that its policy of disparate treatment is necessary to promote a compelling or important governmental interest. See Shapiro v. Thompson, 394 U.S. 618, 634, 89 S.Ct. 1322, 1331, 22 L.Ed.2d 600 (1969); Sherbert v. Verner, 374 U.S. 398, 406, 83 S.Ct. 1790, 1795, 10 L.Ed.2d 965 (1963); Bates v. Little Rock, 361 U.S. 516, 524, 80 S.Ct. 412, 417, 4 L.Ed.2d 480 (1960); Korematsu v. Little Rock, 323 U.S. 214, 216, 65 S.Ct. 193, 194, 89 L.Ed. 194 (1944); Skinner v. Oklahoma, 316 U.S. 535, 541, 62 S.Ct. 1110, 1113, 86 L.Ed. 1655 (1942).
Appellants contend they have made a prima facie showing that the government is engaging in a policy which, on its face, discriminates against those engaging in protected political speech. “[Djetermin-ing whether a prima facie case has been established requires consideration of all relevant circumstances, including whether there has been a pattern of” denial of diversion on the basis of the exercise of constitutional rights. Edmonson v. Leesville Concrete Co., Inc., — U.S. -, 111 S.Ct. 2077, 2088, 114 L.Ed.2d 660 (1991) (citing Batson v. Kentucky, 476 U.S. 79, 96-97, 106 S.Ct. 1712, 1722-1723, 90 L.Ed.2d 69 (1986)). In this case, appellants proffered evidence of two statements: (1) an extrajudicial admission by Ms. Winfree, the chief of the section of the United States Attorney’s office prosecuting appellants’ cases, stating that the United States Attorney had “a policy of not diverting protest cases;” and (2) a statement by a paralegal who conducted diversion interviews in the United States Attorney’s office that the office had a policy not to divert the Farra[382]*382gut West demonstrators.17 These statements were corroborated by the fact that each appellant’s diversion interview was perfunctory and by the powerful statistical evidence that not one of the 275 political demonstrators arrested for unlawful entry had been admitted to the pretrial diversion program, in contrast with 27% of the non-demonstrating unlawful entrants who were granted diversion.18 On these facts, we agree that appellants have made a prima fade showing of a government policy that, by its own terms, more severely punishes those who exercise protected constitutional rights than those who do not.19
Denying diversion to appellants because they are members of the class of political demonstrators is similar to the conduct of city officials who reinstated criminal charges for traffic violations after the defendants filed complaints charging police misconduct. See Dixon v. District of Columbia, 129 U.S.App.D.C. 341, 394 F.2d 966 (1968). In reversing appellant’s conviction, the court wrote:
The Government may not prosecute for the purpose of deterring people from exercising their right to protest official misconduct and petition for redress of grievances. Moveover, a prosecution under such circumstances would be barred by the equal protection clause, since the Government employs an impermissible classification when it punishes those who complain against police misconduct and excuses those who do not.
Id. at 343, 394 F.2d at 968. Similarly, the government — absent an adequate explanation — is not free to punish more severely those unlawful entrants who are political demonstrators, while excusing those who also refuse to leave on demand of lawful authority but are not making a political statement.
In sum, we conclude, on the facts of record, that appellants have made a prima facie showing that the government has a policy which, on its face, impermissibly discriminates against charged offenders based on their exercise of protected First Amendment rights.
D.
We turn to the government’s burden on remand and to the required remand procedures. Appellants’ showing has “shifted the burden of proof to the State to dispel the inference of intentional discrimination.” Castaneda v. Partida, 430 U.S. 482, 497-98, 97 S.Ct. 1272, 1281-82, 51 L.Ed.2d 498 (1977). “Once a prima facie case of invidious discrimination is established, the burden of proof shifts to the State to rebut the presumption of unconstitutional action by showing that permissible ... neutral selection criteria and procedures have produced the monochromatic result.” Alexander v. Louisiana, 405 U.S. 625, 631-32, 92 S.Ct. 1221, 1225-26, 31 [383]*383L.Ed.2d 536 (1972) (citing Turner v. Fouche, 396 U.S. 346, 361, 90 S.Ct. 532, 540, 24 L.Ed.2d 567 (1970); Eubanks v. Louisiana, 356 U.S. 584, 587, 78 S.Ct. 970, 973, 2 L.Ed.2d 991 (1958)). It is important to note that “the prosecutor may not rebut [appellants’] case merely by denying that he [or she] had a discriminatory motive,” Batson, 476 U.S. at 97-98, 106 S.Ct. at 1723-1724; “affirmations of good faith in making individual selections are insufficient to dispel a prima facie case of systematic exclusion.” Alexander, 405 U.S. at 632, 92 S.Ct. at 1226. “The State cannot meet this burden on mere general assertions that its officials did not discriminate or that they properly performed their official duties.” Batson, 476 U.S. at 94, 106 S.Ct. at 1721. “If these general assertions were accepted as rebutting a defendant’s prima facie case, the Equal Protection Clause ‘would be but a vain and illusory requirement.’” Id., at 98, 106 S.Ct. at 1724 (quoting Norris v. Alabama, 294 U.S. 587, 598, 55 S.Ct. 579, 584, 79 L.Ed. 1074 (1935)). Rather, “the prosecutor must give a ‘clear and reasonably specific’ explanation of his [or her] ‘legitimate reasons’ for [denying diversion].” Id. 476 U.S. at 98 n. 20, 106 S.Ct. at 1724 n. 20 (quoting Texas Dep’t of Community Affairs v. Burdine, 450 U.S. 248, 258, 101 S.Ct. 1089, 1096, 67 L.Ed.2d 207 (1981)).
Ordinarily, having made a prima facie showing as to both elements of a selective prosecution claim, appellants would be entitled to proceed immediately with discovery followed by an evidentiary hearing. See supra notes 10, 12. Here, however, a remand order to that effect would be premature. For two reasons, this case is in a posture where the court, on remand, should hold an initial hearing to determine whether the government can give a “clear and reasonably specific” explanation based on “legitimate reasons” for denying diversion that would resolve the case without need for discovery and a full-blown evidentiary hearing. Cf. Greenwood, 796 F.2d at 52, supra note 13; Saade, 652 F.2d at 1135, supra note 14.
(1)
First, the government prevailed before both trial judges on the erroneous ground that appellants were not similarly situated with other charged unlawful entrants. The government was not called upon to respond more than perfunctorily to what we now hold is a prima fade case of improper motivation. To convince the judges, the government merely had to assert, not prove, that appellants had failed to show the government’s enforcement of a content-neutral statute (rather than an offer of diversion) was motivated by a desire to silence protestors on behalf of the homeless. Judge Salzman simply accepted that assertion as the heart of his alternative or secondary ruling, without permitting Fedo-rov’s and Donne’s proffers. Judge Burnett accepted the possibility that all political demonstrators at the Farragut West Metro stop had been prosecuted, not diverted, but he failed to perceive any discriminatory purpose. In neither case, therefore, was the government required — as it is now — to justify its apparent refusal to divert political demonstrators from prosecution while diverting other first-time unlawful entrants.
It is possible (though we have no way of knowing) that the government has a straightforward, simple explanation of its diversion policy that would lead to a prompt resolution of this case without need for discovery. Accordingly, a concern for judicial economy, as well as proper recognition of the government’s broad discretion whether to prosecute, leads us to conclude that before the trial judges20 in these cases order discovery, the government on remand shall have one more opportunity to respond to appellants’ charges, subject to rebuttal. The government, for example, may desire simply to show that appellants’ data are [384]*384incorrect.21 Or the government may decide to offer credible evidence to prove its contention, made for the first time on appeal, that any policy which had the effect of denying diversion to all first-time political demonstrators charged with unlawful entry is justified for a neutral reason unrelated to First Amendment concerns.22 For example, the government has indicated it may attempt to show it had a reasonable desire to discourage crowds that pose a more significant law enforcement problem than other typical unlawful entry cases.23 If the government takes this confession- and-avoidance approach, which poses primarily a question of law, the proceedings on remand are likely to be handled more expeditiously through a limited initial hearing — and less intrusively into the prosecutor’s files — than would be the case if there were an immediate commencement of comprehensive discovery. If, however, upon rebuttal of the government’s case at this initial hearing the defendant-appellants’ prima facie case of selective prosecution survives, then they will be entitled to discovery and a more comprehensive eviden-tiary hearing. We leave it to the trial judge (see supra note 20) to decide how best to proceed if the government takes this approach. We intimate no views on the legitimacy of the government’s position if it does so.
One further point is important here: we leave it to the trial judge’s sound discretion to determine whether the government can sufficiently rebut appellants’ prima facie case on remand merely through sworn affidavits or, in addition, must provide one or more witnesses subject to cross-examination. The answer presumably will depend on the nature of the government’s response {e.g., attack on appellants’ data; justification of policy denying diversion to political demonstrators), as well as on a careful weighing of the government’s interest in avoiding undue interference with prosecu-torial discretion as against appellants’ right to equal protection of the laws in a context where they already have shown a prima fade case of selective prosecution.
(2)
There is a second reason not to order immediate discovery. If the government, in its first response after remand, attempts to rebut appellants’ prima fade case but the evidence it presents does not fully respond to appellants’ case or raises new issues, the trial judge may confront not only appellants’ need for immediate discovery to pursue the constitutional issue, but also the government’s alleged need to protect the confidentiality of some of its files. In that case, the trial judge may have to monitor discovery very carefully by dealing extensively with documents in camera and thus “delvpng] more deeply than [he] might ordinarily into marshalling the evidence on both sides for the selective [385]*385prosecution claim.” Irish People, 221 U.S.App.D.C. at 433, 684 F.2d at 955. The trial judge, as a result, may have to make representative findings of fact from government files in a way that would not compromise the information the government asserts is privileged. See Societe Internationale v. Rogers, 357 U.S. 197, 203, 78 S.Ct. 1087, 1091, 2 L.Ed.2d 1255 (1958); In re Attorney General of the United States, 596 F.2d 58, 67 (2d Cir.), cert. denied, 444 U.S. 903, 100 S.Ct. 217, 62 L.Ed.2d 141 (1979); see, e.g., United States v. Holmes, 794 F.2d 345, 348 n. 3 (8th Cir.1986) (trial court reviewed extensive prosecution records in camera to determine why decision to prosecute had been made in individual cases); United States v. Kahl, 583 F.2d 1351, 1355 (5th Cir.1978) (trial court conducted in camera review to determine if government documents were relevant to selective prosecution claim); Super.Ct.Crim.R. 26.2(c).
In short, our point is that, depending on the government’s position, the trial court on remand should have considerable discretion in shaping the proceedings necessary to address a prima facie case of selective prosecution, keeping in mind both the appellants’ constitutional rights and the prosecutor’s broad discretion over deciding whether to prosecute (or divert from prosecution), including control over confidential government files.
The government has argued — for the first time on appeal — that in all 275 cases of non-diverted political demonstrators it was motivated by concerns about public disruption and use of large numbers of police, not by a desire to chill political protest. Evidence supporting such alleged concerns, however, is not on the record. Because fundamental constitutional rights are implicated in this case, it remains to be seen whether the government can make a convincing showing that meaningfully accounts for the disparate treatment it apparently has accorded political demonstrators charged with unlawful entry when compared with all other similarly situated offenders.
III.
Appellants also contend that the United States Attorney’s policy of denying diversion to all political demonstrators charged with unlawful entry offends the First Amendment. We cannot resolve this issue, however, because the trial judges did not make adequate findings. Because we reverse and remand for further consideration of the selective prosecution claims, we also remand for further findings of fact and conclusions of law on the First Amendment issue in the event that appellants do not prevail on the first issue.
Both trial judges, in denying appellants’ First Amendment claims, merely asserted that no constitutional right was implicated because Metro closed the Farragut West station at a particular time for all purposes, not just for curtailing expressive activities, and that appellants’ failure to receive pretrial diversion, therefore, did not manifest an intended First Amendment infringement. These judicial rationales are not responsive to appellants’ claims for two reasons. First, they implicitly reject, or at least do not account for, the chill on expressive activity reflected in appellants’ prima facie cases of selective prosecution. That prima facie Fifth Amendment violation reflects a denial of equal protection because of discrimination against persons exercising First Amendment rights. Thus, it would appear that, at least presumptively, a First Amendment violation may have occurred as well.
Second, the trial judges did not deal with the proper framework of analysis. It is true that, in United States v. O’Brien, 391 U.S. 367, 376, 88 S.Ct. 1673, 1678, 20 L.Ed.2d 672 (1968), the Supreme Court acknowledged that “when ‘speech’ and ‘non-speech' elements are combined in the same course of conduct” — as is true when a political demonstrator violates the unlawful entry statute by refusing to leave a Metro station at closing time — “a sufficiently important governmental interest in regulating the nonspeech element can justify incidental limitations on First Amendment freedoms.” The court then added that
[386]*386a governmental regulation is sufficiently justified [1] if it is within the constitutional power of the Government; [2] if it furthers an important or substantial governmental interest; [3] if the governmental interest is unrelated to the suppression of free expression; and [4] if the incidental restriction on alleged First Amendment freedoms is no greater than is essential to the furtherance of the interest.
Id. at 377, 88 S.Ct. at 1679. The trial judges never came to grips with these factors. On remand, therefore, the trial court, unless it sustains appellants’ selective prosecution claims, will have to reach the question whether a policy of refusing to divert all political demonstrators from prosecution under the unlawful entry statute “is no greater than essential to the furtherance of,” id., the government’s substantial interest in policing the Farragut West Metro station.
IV.
We reverse and remand these cases for further proceedings in which the government has the burden of rebutting appellants’ prima facie cases of selective prosecution, in violation of the Equal Protection Clause of the Fifth Amendment, and if necessary for further proceedings on appellants’ claimed violations of their First Amendment rights.
Reversed and Remanded.