Plaintiffs, Oregon voters who signed Referendum 308, appeal the district court’s denial of permanent injunctive relief against Oregon Secretary of State Bill Bradbury (“Secretary”). The Secretary determined that Referendum 303, which sought a statewide vote on a legislative act establishing same-sex domestic partnerships, did not have enough valid signatures to qualify for the ballot. Plaintiffs contend that the Secretary’s procedures for verifying referendum petition signatures violated their equal protection and due process rights. The district court held that no constitutional violations occurred. We affirm.

The citizens of Oregon have reserved the power to refer legislative acts to the ballot for approval or rejection by the state’s voters. Or. Const, art. IV, § 1(3). A referendum qualifies for a statewide vote upon submission of a petition containing valid signatures of “a number of qualified voters equal to four percent of the total number of votes cast for all candidates for Governor” in the most recent gubernatorial election. Id. § l(3)(b). The deadline for submission of a referendum petition is ninety days after the end of the legislative session in which the legislation was passed. Id. The Secretary must complete the signature verification process within thirty days of the petition filing deadline. Id. § l(4)(a).

The Oregon Constitution subjects the people’s referendum power to regulation by the legislature. See id. § l(4)(a)-(b). Statutes require the Secretary to use a statistical sampling method for verifying referendum petition signatures, using “an elector’s voter registration record or other database.” Or.Rev.Stat. § 250.105(4), (6). The Secretary samples approximately five percent of the submitted signatures for each referendum, and submits the sampled signatures to county elections officials. Or. Admin. R. 165 -014-0030(5)(a), (7). The State Initiative and Referendum Manual, adopted by the Secretary through administrative rule, see Or. Admin. R. 165 -014-0005(1), requires county elections officials to “verify[ ] the original signatures” sampled from referendum petitions “using voter registration records.” A publication entitled Directive for Signature Verification, issued by the Secretary on November 24, 1981, specifies that county elections officials should “[cjompare the signature on the petition and the signature on the voter registration card to identify whether the signature is genuine and must be counted.” The Secretary extrapolates the overall number of valid petition signatures using the sampled signature results. See Or. Admin. R. 165 -014-0030.

Under these procedures, county elections officials verify sampled referendum signatures by determining whether each petition signature matches the signature on the signer’s existing voter registration card. Oregon law does not provide procedures by which a voter can introduce extrinsic evidence to rehabilitate a referendum signature after its rejection. See Or. Rev.Stat. § 250.105. In contrast, for non-matching ballot signatures returned during Oregon’s vote-by-mail elections, the Secretary’s Vote by Mail Procedures Manual requires county elections officials to give the voter ten-days notice and an opportunity to submit an updated voter registration card.

In 2007, the Oregon Legislature passed House Bill 2007 (“HB 2007”), which establishes same-sex domestic partnerships. See 2007 Or. Laws, ch. 99. Plaintiffs are Oregon voters who signed petitions for Referendum 303, which sought a statewide vote on HB 2007. On September 26, 2007, proponents of Referendum 303 submitted approximately 62,000 unverified signatures to the Secretary. The Secretary randomly selected 3,033 signatures from the petition through statistical sampling.

On October 3, 2007, the Secretary distributed the sampled signatures to county elections officials. An accompanying memorandum instructed county officials to “[cjheck the registration of the names indicated by comparing the signature on the petition to the signature on the registration card.” One of the options for rejecting signatures was that the “[signatures do not match.” The memorandum instructed the counties to complete their verification by October 8, 2007.

Counties allowed petitioners and members of the public to observe the signature verification process. Every county had a system for reviewing initially rejected signatures. The counties rejected 254 of the 3,033 sampled signatures; of the rejected signatures, 55 were invalid because they did not match signatures on existing voter registration cards. No county gave notice to voters with rejected signatures. The counties also refused to consider extrinsic evidence presented by voters with rejected signatures, such as affidavits and updated voter registration cards.

On October 26, 2007, after extrapolating the verification results, the Secretary announced that there were 55,083 valid signatures for Referendum 303, fewer than the 55,179 valid signatures required to qualify for the ballot. Plaintiffs filed an action in federal district court against the Secretary and several county officials, alleging violations of state and federal rights, including equal protection and due process. Plaintiffs sought an opportunity to rehabilitate their signatures, and a declaration that Referendum 303 contained enough signatures to qualify for the ballot. On December 31, 2007, the district court granted plaintiffs’ motion for a preliminary injunction, staying the effective date of HB 2007 until a hearing on the merits.

After hearing oral argument on February 1, 2008, the district court vacated the preliminary injunction, denied permanent injunctive relief, and dismissed plaintiffs’ federal claims. First, the court held that the Secretary’s signature verification procedures did not violate plaintiffs’ equal protection rights. The court concluded that signing a referendum petition does not implicate a fundamental right, and in any event the Secretary’s procedures do not unconstitutionally infringe on plaintiffs’ alleged rights. The court also ruled that county elections officials use specific, uniform standards for signature verification. Second, the district court held that the Secretary did not violate plaintiffs’ due process rights by denying them notice and an opportunity to rehabilitate their rejected referendum signatures.

This appeal followed.

III

We review the district court’s denial of permanent injunctive relief for abuse of discretion. Citizens for Clean Gov’t v. City of San Diego, 474 F.3d 647, 650 (9th Cir.2007). Legal conclusions underlying the denial are reviewed de novo. Id. We review factual determinations for clear error. Scott v. Pasadena Unified Sch. Dist., 306 F.3d 646 , 653 (9th Cir.2002).

The Fourteenth Amendment provides that no state shall “deprive any person of life, liberty, or property, without due process of law; nor deny to any person within its jurisdiction the equal protection of the laws.” U.S. Const, amend. XIV, § 1. Plaintiffs do not challenge the sampling concept, but contend that the Secretary’s procedures for verifying sampled signatures violate their equal protection and due process rights. First, they argue that the method of verifying referendum signatures burdens their fundamental right to vote. Second, plaintiffs contend that the Secretary violated their equal protection and due process rights by not providing the same notice and opportunity to rehabilitate referendum signatures that is afforded to signers of vote-by-mail election ballots. Third, plaintiffs argue that county elections officials lack uniform statewide rules for verifying referendum signatures, violating the rule from Bush v. Gore, 531 U.S. 98 , 121 S.Ct. 525 , 148 L.Ed.2d 388 (2000).

We hold that the Secretary’s procedures did not violate plaintiffs’ constitutional rights. Although regulations on the referendum process implicate the fundamental right to vote, the state’s important interests justify the minimal burden on plaintiffs’ rights. Additionally, Oregon’s signature verification standards are uniform and specific enough to ensure equal treatment of voters.

We agree with plaintiffs that state regulations on the initiative and referendum process implicate the fundamental right to vote. In Moore v. Ogilvie, 394 U.S. 814 , 89 S.Ct. 1493 , 23 L.Ed.2d 1 (1969), the Supreme Court held that restrictions on candidate nominating petitions implicate voters’ fundamental rights because “[a]ll procedures used by a State as an integral part of the election process must pass muster against the charges of discrimination or of abridgement of the right to vote.” Id. at 818 , 89 S.Ct. 1493 . In Idaho Coalition United for Bears v. Cenarussa, 342 F.3d 1073 (9th Cir.2003), we relied on Moore and held that regulations on a state’s initiative process “implicate the fundamental right to vote, for the same reasons and in the same manner,” as regulations on candidate nominating petitions. Id. at 1077. “[Wjhile a state may decline to grant a right to legislate through ballot initiatives, it may not grant that right on a discriminatory basis.” Id. at 1077 n. 7.

The principle underlying Moore and Idaho Coalition applies equally to Oregon’s regulations on the referendum pro cess. The people of Oregon reserved both the initiative and referendum powers in their state constitution. Or. Const, art. IV, § l(2)-(3). Both processes serve as “ ‘basic instruments] of democratic government.’ ” Idaho Coalition, 342 F.3d at 1076 (quoting City of Cuyahoga Falls v. Buckeye Cmty. Hope Found., 538 U.S. 188, 196 , 123 S.Ct. 1389 , 155 L.Ed.2d 349 (2003)). Thus, regulations on Oregon’s referendum process implicate plaintiffs’ fundamental right to vote.

Plaintiffs argue that because the Secretary’s referendum verification procedures implicate the right to vote, strict scrutiny applies, and the regulations must be narrowly drawn to advance a compelling state interest. They contend that the procedures cannot meet strict scrutiny because they treat referendum signatures differently than signatures on vote-by-mail ballots. Specifically, plaintiffs note that Oregon limits its referendum verification process to a comparison of petition signatures and signatures on existing voter registration cards. County elections officials do not notify voters after rejecting referendum signatures as non-matching. In contrast, when county elections officials reject a signature on a vote-by-mail election ballot, they give the voter ten-days notice and an opportunity to submit an updated voter registration card, or otherwise provide proof that the signature is valid. Plaintiffs contend that this disparate treatment of referendum petition signatures and vote-by-mail ballot signatures violates their equal protection and due process rights.

Plaintiffs’ argument “proceeds from the erroneous assumption that a law that imposes any burden upon the right to vote must be subject to strict scrutiny.” Burdick v. Takushi, 504 U.S. 428, 432 , 112 S.Ct. 2059 , 119 L.Ed.2d 245 (1992). This premise is flawed. “States retain the power to regulate their own elections,” and “ ‘as a practical matter, there must be a substantial regulation of elections if they are to be fair and honest and if some sort of order, rather than chaos, is to accompany the democratic processes.’ ” Id. at 433 , 112 S.Ct. 2059 (quoting Storer v. Brown, 415 U.S. 724, 730 , 94 S.Ct. 1274 , 39 L.Ed.2d 714 (1974)). Subjecting every election law to strict scrutiny “would tie the hands of States seeking to assure that elections are operated equitably and efficiently.” Id.; see also Crawford v. Marion County Election Bd., — U.S. -, 128 S.Ct. 1610, 1616 , 170 L.Ed.2d 574 (2008).

In Burdick , the Supreme Court recognized that “a more flexible standard applies” for analyzing election laws that burden the right to vote. Id. at 434 , 112 S.Ct. 2059 .

Id. (internal quotations and citations omitted). Under this standard, strict scrutiny applies only when the right to vote is “subjected to severe restrictions.” Id. (internal quotations omitted). However, “when a state election law provision imposes only reasonable, nondiscriminatory restrictions upon the First and Fourteenth Amendment rights of voters, the State’s important regulatory interests are generally sufficient to justify the restrictions.” Id. (internal quotations omitted).

To date, the Supreme Court has subjected only two types of voting regulations to strict scrutiny. See Green v. City of Tucson, 340 F.3d 891, 899-900 (9th Cir.2003). First, strict scrutiny applies to “regulations that unreasonably deprive some residents in a geographically. defined governmental unit from voting in a unit wide election.” Id. at 899 . Examples include laws that condition the right to vote on property ownership or payment of a poll tax. See id. (citing cases). Second, “regulations that contravene the principle of ‘one person, one vote’ by diluting the voting power of some qualified voters within the electoral unit” also are subject to strict scrutiny. Id. at 900 . Examples include laws that weigh votes from rural counties more heavily than votes from urban counties. See id. (citing cases).

Unlike these examples of severe restrictions, the Secretary’s procedures in this case are reasonable, and do not violate plaintiffs’ equal protection rights. First, the magnitude of plaintiffs’ asserted injury is minimal. See Burdick, 504 U.S. at 434 , 112 S.Ct. 2059 . Although county elections officials do not notify voters after rejecting non-matching signatures, referendum petition cover sheets instruct voters to “[s]ign your full name, as you did when you registered to vote.” Chief petitioners and members of the public observe the process and can object to signature verification decisions. All counties provide that higher county elections authorities review all signatures that are initially rejected. All counties also uniformly limit their review to a comparison between petition signatures and existing voter registration cards. This process protects the rights of petition signers and treats voters in different counties equally. When viewed in context, the absence of notice and an opportunity to rehabilitate rejected signatures imposes only a minimal burden on plaintiffs’ rights.

Second, Oregon’s important interests justify this minimal burden on the right to vote. See id. Oregon elections officials may process more than 100,000 sampled initiative and referendum signatures within the thirty-day period required by state law. In any election period, there may be ten or more proposed initiative and referendum measures that require signature verification. The administrative burden of verifying a referendum petition signature is significantly greater than the burden associated with verifying a vote-by-mail election ballot signature. Verification of a vote-by-mail signature takes mere seconds because elections officials can scan the bar-code on the back of the ballot envelope and automatically access the signer’s voter registration record. In contrast, verification of each referendum petition signature takes several minutes because elections officials must identify the signer, find the corresponding voter registration card, determine whether the signer is an active, registered voter, and then compare the signatures. Moreover, fraudulent signatures are less likely in vote-by-mail elections, in which the ballots are sent directly by the elections official to the voter, and returned directly by the voter to the elections official. In initiative and referenda, by contrast, the signatures are often gathered by privately hired signature gatherers who are paid a fixed amount for each signature they obtain. These differences between referendum petitions and vote-by-mail ballots justify the minimal burden imposed on plaintiffs’ rights in this case.

We reject plaintiffs’ procedural due process argument for the same reasons. Oregon’s interests in detecting fraud and in the orderly administration of elections are weighty and undeniable. Requiring the state to provide thousands of petition signers with individual notice that then-signatures have been rejected and to af ford them an opportunity to present extrinsic evidence during the short thirty-day verification period would impose a significant burden on the Secretary and county elections officials. In contrast to the significant weight of the state’s interests, plaintiffs’ interest in the additional procedures they seek is slight. First, the verification process is already weighted in favor of accepting questionable signatures, in part because only rejected signatures are subject to more than one level of review by county elections officials. Providing notice and allowing individuals to contest a determination that a signature did not match would further skew the process in favor of accepting invalid signatures, as there would be no corresponding notice to those whose signatures were erroneously deemed to match. Second, as previously noted, the Secretary’s procedures already allow chief petitioners and members of the public to observe the signature verification process and challenge decisions by county elections officials. The value of additional procedural safeguards therefore is negligible, and the burden on plaintiffs’ interests from the state’s failure to adopt their proposed procedures is slight at most. See Protect Marriage Ill. v. Orr, 463 F.3d 604, 608 (7th Cir.2006) (holding that “[t]he cost of allowing tens of thousands of people to demand a hearing on the validity of their signatures would be disproportionate to the benefits” of allowing petition signers to challenge rejected signatures).

Under the circumstances, the administrative burden of the additional process plaintiffs propose outweighs any marginal benefit that would result from additional procedures. Thus, Oregon’s “important regulatory interests” are sufficient to justify the state’s “reasonable, nondiscriminatory restrictions.” Burdick, 504 U.S. at 434 , 112 S.Ct. 2059 . Like other “evenhanded restrictions that protect the integrity and reliability of the electoral process itself,” the state’s signature verification procedures must be upheld. Crawford, 128 S.Ct. at 1616 ; Burdick, 504 U.S. at 434 , 112 S.Ct. 2059 .

Plaintiffs also contend that Oregon lacks uniform statewide rules for verifying referendum petition signatures, violating equal protection rights under Bush v. Gore, 531 U.S. 98 , 121 S.Ct. 525 , 148 L.Ed.2d 388 (2000). The Secretary argues that county elections officials use a uniform standard: whether a referendum petition signature matches the signature on the signer’s existing voter registration card. The district court held that this standard is uniform and specific enough to ensure equal treatment of voters. We agree.

In Bush , the Supreme Court reversed the Florida Supreme Court’s decision ordering a manual recount in the disputed 2000 Presidential election. 531 U.S. at 102-03, 111 , 121 S.Ct. 525 . The Florida Supreme Court had instructed county elections officials to count all votes in which there was a “clear indication of the intent of the voter.” Id. at 102 , 121 S.Ct. 525 . In a ruling “limited to the present circumstances,” and applicable only because it was a court-ordered recount, the Supreme Court held that Florida’s recount violated the Equal Protection Clause because it lacked “specific standards to ensure its equal application,” resulting in the disparate treatment of voters, and a process without “sufficient guarantees of equal treatment.” Id. at 106-07, 109 , 121 S.Ct. 525 .

Even were Bush applicable to more than the one election to which the Court appears to have limited it, Oregon’s standard for verifying referendum signatures would be sufficiently uniform and specific to ensure equal treatment of voters. The Secretary uniformly instructs county elections officials to verify referendum signatures by determining whether each petition signature matches the signature on the signer’s voter registration card. For example, the Secretary’s November 24, 1981 Directive for Signature Verification instructs county elections officials to “Compare the signature on the petition and the signature on the voter registration card to identify whether the signature is genuine and must be counted.” Similarly, the Secretary’s October 3, 2007 memorandum accompanying sampled Referendum 303 signatures instructed county officials to “[c]heck the registration of the names indicated by comparing the signature on the petition to the signature on the registration card.” One of the options provided for rejecting Referendum 303 signatures was that the “[signatures do not match.” This standard ensures equal treatment because it uniformly requires all counties to compare two existing signature specimens when determining the validity of each sampled signature on a referendum petition. Moreover, the task is far simpler than that used in many states to determine voter intent. See id. at 106-07 , 121 S.Ct. 525 .

Oregon’s referendum signature verification process also has “sufficient guarantees of equal treatment.” Id. at 107 , 121 S.Ct. 525 . The Secretary sponsors signature verification training sessions, and county elections officials regularly attend these sessions and use the materials provided. During the verification of Referendum 303, all counties subjected initially rejected signatures to a second level of review. The signature verification process was uniformly limited to a comparison between petition signatures and existing voter registration cards, and all counties refused to consider extrinsic" evidence from rejected signers.

We are not persuaded by plaintiffs’ arguments to the contrary. Plaintiffs contend that a few petitioners had signatures rejected on Referendum 303, even though their signatures were accepted on petitions for other referenda. Such isolated discrepancies do not demonstrate the absence of a uniform standard, nor do they prove that the district court was clearly erroneous in finding that “overall we have a fairly coherent set of results.” After all, the fact that the same name appeared on both petitions does not establish that both signatures matched the voter registration card. Plaintiffs also argue that county elections officials exercised discretion when deciding whether to review erroneously excluded signatures. They point to Washington County’s decision to amend its verification results, accepting one signature that was initially rejected. As the record shows, however, Washington County properly based its revised decision on a comparison of the voter’s petition signature and existing voter registration card. Like all other counties, Washington County did not review extrinsic evidence.

Finally, plaintiffs point to differences in the number of signatures rejected by various counties. For example, Hood River County rejected three of its twenty-one sampled signatures as non-matching; Multnomah County excluded only one non-matching signature from its 274 sampled signatures; and Washington County rejected seventeen of its 336 sampled signatures as non-matching. Plaintiffs contend that these differences demonstrate the absence of a uniform standard. This argument is without merit, for the reasons given by the district court. First, the statistical significance of these differences is questionable, considering the relatively small number of sampled signatures at issue. Second, signature gatherers in some counties do a better job than those in other counties. Most importantly, uniform standards can produce different results. Viewing the record as a whole, we are convinced that Oregon’s counties used the same uniform standard and thus would satisfy the requirements of Bush , if it were applicable.

For these reasons, we hold that the Secretary’s procedures for verifying sampled referendum petition signatures did not violate plaintiffs’ equal protection and due process rights.

AFFIRMED.

. Other reasons included "[n]ot registered” and "[illegible signatures.” . The district court also dismissed plaintiffs' claim that the Secretary's actions violated the First Amendment by prohibiting core political speech. Plaintiffs do not appeal the district court's ruling on this claim. . The state’s expert reviewed 556 of the signatures county elections officials verified as matching, only a small percentage of the total number of Referendum 303 signatures. She found sixty-five signatures in that small portion in which she would have ruled differently than the county officials — i.e., in sixty-five cases the handwriting expert would have concluded that the accepted petition signature did not match the voter registration card. By contrast, the expert reviewed all of the signatures that county elections officials determined did not match, and found only six cases in which she would have ruled differently. Thus, if the handwriting expert had verified all of the signatures, the petition would have lacked sufficient signatures by a far wider margin than the county elections officials determined.