III. Discussion
A. Standard of Review
The district court denied Defendant's motion to dismiss based on its conclusion Plaintiffs were entitled to judgment on the pleadings as a matter of law. We review this ruling de novo. Utah Republican Party v. Cox , 892 F.3d 1066, 1076 (10th Cir. 2018).
B. Fourteenth Amendment Claim
Because this matter was decided on the pleadings, this court turns first to the allegations in Plaintiffs' complaint. There is, of course, no dispute that Section 2.5 requires proponents of ballot initiatives to collect signatures from two percent of the registered voters in each of Colorado's state senate districts. Plaintiffs' complaint alleges that the population of each district varies and, thus, Section 2.5 "dilutes the value of the signature of voters in densely populated senate districts and gives them less value than the signatures of voters in sparsely populated districts."5 If true, this allegation may support the grant of judgment in favor of Plaintiffs. See Reynolds v. Sims , 377 U.S. 533, 568, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964) (holding that "the Equal Protection Clause requires that the seats in both houses of a bicameral state legislature must be apportioned on a population basis"). The allegation, however, is not true. The Colorado state senate districts are equally populous and Plaintiffs concede this point in their appellate brief. Appellee Br. at 2 (admitting that Colorado's state senate "districts are approximately equal in total population"); see also Evenwel v. Abbott , --- U.S. ----, 136 S. Ct. 1120, 1124, 194 L.Ed.2d 291 (2016) (stating that "all States use total-population numbers from the census when designing congressional and state-legislative districts"). Thus, the allegation cannot be used to support the district court's ruling in favor of Plaintiffs on their equal protection claim.
Plaintiffs' complaint, however, also alleges that the number of registered voters in each state senate district differs considerably. Specifically, it states:
There is a huge variation in the population of registered voters in the various state senate districts. For example, as of January 1, 2017, district 11 had 86,181 voters, district 25 had 85,051 voters, district 21 had 80,499 voters, and five other districts (1, 12, 13, 29, and 35) had between 91,728 and 96,463 voters. By way of comparison, district 4 had 121,093 voters, district 16 had 119,920 voters, district 18 had 120,222 voters, district 20 had 126,844 voters, and district 23 had 132,222 voters. Thus, district 23 has 51,723 more voters than district 21, and that variance is slightly more than 60%.
Presuming this allegation to be true,6 the complaint can be read to allege that an inequality in the number of registered voters *1139in each of Colorado's equally populous senate districts dilutes the voting rights of petition signatories who live in districts with a higher number of registered voters.7 See Evenwel , 136 S. Ct. at 1125 (involving the same assertion by voters in Texas). In other words, because only the signatures of registered voters are valid for purposes of citizen-initiative petitions, the number of signatures required to meet the two-percent threshold established by Section 2.5 varies from district-to-district.
Using the numbers alleged by Plaintiffs, approximately 1610 signatures must be collected in District 21 to satisfy the two percent requirement, but 2537 signatures must be collected in District 20 (the district in which Plaintiff Hayes resides) and 2404 signatures must be collected in District 18 (the district in which Plaintiff Semple resides). As the argument goes, Plaintiffs' votes have less influence on whether a citizen initiative appears on the state-wide ballot than the votes of individuals living in districts with fewer registered voters. Defendant understood this to be Plaintiffs' assertion and addressed it in her motion to dismiss. She argued Plaintiffs' claim fails as a matter of law because every court to consider the matter has held that signature-collection requirements involving ballot initiatives do not violate the Equal Protection Clause as long as the districts from which signatures are collected have substantially the same total population. See Angle v. Miller , 673 F.3d 1122, 1131 (9th Cir. 2012) (upholding Nevada law requiring signatures from ten percent of registered voters in each equally populous congressional district); Libertarian Party of Va. v. Davis , 766 F.2d 865, 868 (4th Cir. 1985) (upholding requirement of 200 signatures from each of Virginia's ten congressional districts because the districts "contain, as nearly as practicable, an equal number of inhabitants"); Libertarian Party v. Bond , 764 F.2d 538, 539, 544 (8th Cir. 1985) (upholding Missouri's "one percent in each" or a "two percent in one-half" signature requirement because the congressional districts were "virtually equal in population").
Recognizing that the cases on which she relied did not involve allegations that equally populous districts had unequal numbers of registered voters, Defendant further argued the cases were nonetheless applicable because the Supreme Court recently held in Evenwel v. Abbott that the Equal Protection Clause does not require states to draw their legislative districts based on registered-voter population rather than total population even if the two numbers differ. 136 S. Ct. at 1132-33. Defendant argued the Court's reasoning in Evenwel applied to Plaintiffs' Fourteenth Amendment claim. In their response, Plaintiffs argued Evenwel was inapposite because it involved legislative apportionment and equality of representation while their complaint implicated only ballot-access issues. The district court agreed with Plaintiffs, concluding Evenwel involved "the tension between preventing vote dilution and ensuring equality of representation," a tension it stated was not present here because this matter only involved allegations of vote dilution, not equality of representation. Semple , 290 F. Supp. 3d at 1197. This was error.8
*1140Evenwel involved a challenge to the practice of drawing state legislative districts based on total population rather than voter-eligible population. 136 S. Ct. at 1123. The plaintiffs, two Texas voters, lived in state senate districts "with particularly large eligible- and registered-voter populations." Id . at 1125. They argued that basing apportionment on total population, rather than voter population, unconstitutionally diluted their votes in relation to voters in other senate districts. Id . The Court rejected plaintiffs' argument, holding that longstanding precedent allowed states to draw legislative districts based on total population. Id . at 1123. It reasoned that "[a]dopting voter-eligible apportionment as constitutional command would upset a well-functioning approach to districting that all 50 States and countless local jurisdictions have followed for decades, even centuries."9 Id . at 1132.
Although Evenwel involved the right to vote, not the right to sign a ballot initiative petition, its reasoning governs the outcome of Plaintiffs' equal protection claim. Section 2.5 requires that signatures be obtained from a sub-group of the total population in each state legislative district. Just as in Evenwel , Plaintiffs allege this sub-group varies in size from district to district. Thus, Plaintiffs' equal protection claim is the same as that of the plaintiffs in Evenwel -that the population of either eligible or registered voters, not the total population, in each state senate district must be equal or voting power is diluted. The Court rejected this proposition, refusing to hold that the principle of one person, one vote requires states to equalize the number of voters in each legislative district. Id . at 1131 ("It would hardly make sense for the Court to have mandated voter equality sub silentio and then used a total-population baseline to evaluate compliance with that rule.").
Notwithstanding the Court's statements in Evenwel , Plaintiffs argue Evenwel 's holding is inapplicable in this matter because it was based on an analysis *1141that balanced representational equality against vote dilution. The district court agreed, concluding Section 2.5 does not involve legislative apportionment and, thus, there is no representational equality to balance against the alleged vote dilution. Semple , 290 F. Supp. 3d at 1197-98. Plaintiffs, however, rightly conceded during oral argument in this matter that citizen initiatives and direct democracy do, in fact, implicate the principle of representational equality. As the Court stated in Evenwel , "[n]onvoters have an important stake in many policy debates-children, their parents, even their grandparents, for example, have a stake in a strong public-education system-and in receiving constituent services, such as help navigating public-benefits bureaucracies." 136 S. Ct. at 1132. Because elected representatives make decisions that affect both voting and nonvoting constituents, representation is equal when total population in each district is equal. Id . at 1130-31. The same is true in the direct democracy context. Because legislators "serve all residents, not just those eligible or registered to vote,"10 voters choose a legislator who they expect will adequately advance the interests of both voters and non-voters. In the direct democracy context, voters are able to directly advance the interests of non-voting members of their families and communities when they decide whether to support a citizen initiative. Although citizens, unlike elected representatives, are not "subject to requests and suggestions"11 from constituents, their vote on citizen initiative petitions can be influenced by private discussions with non-voting friends, family, and neighbors. In this way, voting-eligible citizens assume a role similar to that of elected representatives when those voters engage in the initiative process.12 Otherwise, the interests of nonvoters, who the Court acknowledges "have an important stake in many policy debates,"13 would be stifled in the direct-democracy process. This interpretation is consistent with the Supreme Court's statement in Evenwel that the "one-person, one-vote guarantee" can properly be viewed as a guarantee of equal representation, "not voter equality." Id . at 1131.
Supreme Court precedent is clear. No equal protection problem exists if votes are cast in equally populated state legislative districts that were drawn based on Census population data. Id . In no instance has the Court "determined the permissibility of [perfect population] deviation based on eligible- or registered-voter data." Id . Just as it is not unconstitutional to apportion seats in a state legislature based on districts of equal total population, id. at 1124, it is not unconstitutional to base direct democracy signature requirements on total population.14 The district *1142court's basis for distinguishing Evenwel is, thus, unconvincing and led the court to erroneously grant judgment in favor of Plaintiffs on their equal protection claim. Evenwel controls the disposition of Plaintiffs' equal protection claim. Because there is no dispute that Colorado's thirty-five state senate districts are approximately equal in total population, summary judgment must be entered in favor of Defendant on that claim.
C. First Amendment Claims
Plaintiffs raise two First Amendment challenges to Section 2.5, asserting it violates the First Amendment by (1) increasing the cost and difficulty of placing an initiative measure on the ballot and (2) compelling core political speech in some senate districts.
As to their first theory, Plaintiffs complain that Section 2.5 unduly burdens their First Amendment rights by making it more difficult to place a citizen initiative on the state-wide ballot. Specifically, Plaintiffs' complaint alleges the following in the third claim for relief:
By requiring initiative proponents to gather signatures from each of the state's thirty-five senate districts, Amendment 71 significantly increases the cost and difficulty of placing an initiated constitutional amendment on the general election ballot because it is far more efficient and far more cost effective for circulators to collect signatures in densely populated senate districts than it is for them to collect signatures in rural districts where the population density is very low.
Even assuming Plaintiffs are able to prove the relevant factual allegations in their complaint, this court has previously addressed and rejected the proposition that the First Amendment is implicated by a state law that makes it more difficult to pass a ballot initiative. Initiative & Referendum Instit. v. Walker , 450 F.3d 1082, 1099-1103 (10th Cir. 2006) (en banc). In Walker , the plaintiffs argued that a supermajority requirement for passage of wildlife initiatives in Utah impermissibly burdened the exercise of their First Amendment rights by making such initiatives less likely to succeed. Id . at 1098. This court, sitting en banc, acknowledged that the First Amendment protects political speech incident to an initiative campaign but held that the supermajority requirement at issue did not violate the First Amendment. Id . at 1099-1103. Relying on this court's holding in Save Palisade FruitLands v. Todd , 279 F.3d 1204 (10th Cir. 2002), the Walker court emphasized the distinction "between laws that regulate or restrict the communicative conduct of persons advocating a position in a referendum, which warrant strict scrutiny, and laws that determine the process by which legislation is enacted, which do not." Walker , 450 F.3d at 1099-1100. In this way, the Utah law at issue in Walker differed from the Colorado law at issue in Meyer v. Grant , 486 U.S. 414, 108 S.Ct. 1886, 100 L.Ed.2d 425 (1988), which "dictated who could speak ... or how to go about speaking." Walker, 450 F.3d at 1099.
Walker controls the issue presented here because Section 2.5 merely determines the process by which initiative legislation is enacted in Colorado. Like the supermajority requirement addressed by the en banc court in Walker , Section 2.5 is not content-based. Thus, even assuming Section 2.5 makes it more difficult and costly to amend the Colorado constitution because it requires Plaintiffs to collect signatures from all districts in the state, that process requirement does not give rise to a cognizable First Amendment claim. Because Plaintiffs' first theory fails as a matter of law, judgment must be entered in favor of Defendant on this claim.
*1143Plaintiffs also argue Section 2.5 violates the First Amendment by requiring them to interact in certain districts they prefer to avoid. Their complaint alleges Section 2.5 compels core political speech "[b]y requiring initiative proponents to gather signatures from voters in every state senate district, and to engage in political speech and associational activities in each of those thirty-five senate districts ... even though, in the absence of Amendment 71's requirements, they would avoid engaging in political speech and associational activities in those districts."
As a general matter, the First Amendment protects an individual's "right to speak freely and the right to refrain from speaking at all." Wooley v. Maynard , 430 U.S. 705, 714, 97 S.Ct. 1428, 51 L.Ed.2d 752 (1977). To state a compelled-speech claim, a plaintiff must establish three elements: (1) speech; (2) to which the speaker objects; that is (3) compelled by some governmental action. Cressman v. Thompson , 798 F.3d 938, 951 (10th Cir. 2015). As to the compulsion element, this court has held that "the governmental measure must punish, or threaten to punish, protected speech by governmental action that is regulatory, proscriptive, or compulsory in nature." Axson-Flynn v. Johnson , 356 F.3d 1277, 1290 (10th Cir. 2004) (quotation omitted). A plaintiff can show government compulsion without identifying a direct threat, "such as imprisonment, fines, injunctions or taxes," id . (quotations omitted), but a discouragement that is "minimal" and "wholly subjective," does not impermissibly deter the exercise of a plaintiff's First Amendment rights. Phelan v. Laramie Cty. Cmty. Coll. Bd. of Trs. , 235 F.3d 1243, 1247, 1248 (10th Cir. 2000) (quotations omitted) ("Examples of governmentally imposed injuries include: denial of state bar admission, loss of employment, and the conditioning of employment on a vague oath." (citations omitted)).
According to Plaintiffs, they are compelled by Section 2.5 to interact with voters in all state senate districts and if they fail to do so their proposed initiative will not appear on the state-wide ballot.15 Thus, the consequence Plaintiffs allegedly face for failing to comply with Section 2.5 is the failure of their ballot initiative. Plaintiffs have not identified, nor could we find, any precedent holding that the failure of a ballot initiative is an adverse government action that discourages or penalizes the exercise of First Amendment rights. This is not surprising because, taken to its logical end, Plaintiffs' approach would embroil the federal courts in nearly every procedural hurdle imposed by state legislatures on the citizen initiative process. And, as we discussed above, "laws that determine the process by which legislation is enacted" do not implicate the First Amendment. Walker , 450 F.3d at 1100.
The communication of the ideas and beliefs underlying a proposed initiative is not dependent on whether the initiative ultimately appears on the state-wide ballot. Section 2.5 does not erect any barrier to the expression of those ideas and beliefs and Plaintiffs have not argued they have a constitutionally protected right to express themselves through a ballot initiative. Thus, we conclude the consequence of which Plaintiffs complain is not the type of state-mandated penalty necessary to establish a compelled speech claim because that consequence has only a minimal impact on Plaintiffs' First Amendment rights.
IV. Conclusion
The district court's order granting judgment in favor of Plaintiffs is reversed . The *1144matter is remanded to the district court with instructions to grant judgment in favor of Defendant on all claims raised in Plaintiffs' complaint.