This is a case of wait-and-hurry-up, rather than its. more familiar cousin, hurry-up-and-wait. With just two and a half months before the November 2016 general election, Patrick Harlan, the Republican Party’s candidate for an Illinois congressional- seat, and the Crawford County (IL) Republican Central Committee, filed this lawsuit and promptly sought a preliminary ■injunction against the implementation of a state law that allows voters to register and vote on Election Day itself. Generally speaking, the law gives more options for same-day registration and voting for resi*756dents of counties with populations of 100⅛-000 or more than it does for those who live in smaller counties. The plaintiffs contended that the . difference violated their rights under the Fourteenth Amendment’s Equal Protection Clause. The district court agreed with them and issued the injunction; this court granted a stay of that injunction. We now vacate the preliminary injunction altogether.
I
The law .that inspired this litigation is one in a series of measures the Illinois General Assembly has enacted in recent years to facilitate the process of voting for the state’s citizens. Traditionally, Illinois’s deadline for registering to vote was 28 days before the election in question. 10 ILCS 5/5-5. Over the years, the General Assembly added such options as early voting and an extended registration “grace period,” and in 2014 it experimented with a pilot program for election-day registration. See P.A. 98-691. On January 12, 2015, the legislature enacted P.A. ’ 98-1171, which took effect on June 1 of that year; the new law again expanded voting opportunities. It extended the registration grace period, which previously had ended on the third day before an election, to include Election Day, resulting in the following system:
During this grace period [within 28 days of the election], an unregistered qualified elector may register to vote, and a registered voter may submit a change of address form, in person at the office of the election authority, at a permanent polling place established under section 19A-10, at any other early voting site beginning 15 days prior to the election, at a polling place on election day, or at a voter registration location specifically designed for this purpose by the election authority.
10 ILCS 5/4-50, ¶ 1 (emphasis added). The law also permits the voter to cast his or her ballot at the place of registration, even if that is not the person’s normal precinct. Id. ¶ 2.
Recognizing that this expansion of options might be a burden on smaller counties, the legislature added an opt-out provision for counties with populations under 100,000. (“smaller counties”). See 10 ILCS 5/4-50, ¶ 5. Elections in Illinois are administered by county clerks, and in some cities, by Boards of Election Commissioners; the State Board of Elections supervises the overall process. 10 ILCS 5/lA-8(12). If the county maintains an electronic version of the- paper pollbook (called an “e-poll-book”), the 2015 law requires it to offer election-day registration at all precinct polling places no matter what the county’s size. Smaller counties that do not have e-pollbooks are not required under the legislation to offer election-day registration at all polling places, although they may include that option if they wish. But even if they opt not to do so, they must still offer election-day registration at “the election authority’s main office,” as well as at “a polling place in each municipality where 20% or more of the county’s residents reside if the election authority’s main office is not located in that municipality.” 10 ILCS 5/4-50, ¶ 5(i) and (ii). The election authority may also establish other grace periods and voting sites on election days, so long as it gives proper notice of its action. Id. ¶ 5.
Illinois has 102 counties. Twenty of those counties have populations of 100,000 or more; those 20 counties are home to approximately 84% of the state’s residents. See Illinois Population Estimates by County, United States Census Bureau, https:// factfinder.census.gov/bkmk/table/l.O/en/ PEP/2016/PEPANNRES/0400000US 17%C00400000US17.05000. - In addition, Grundy County, Brown County, Stark County, and the city of Danville offered election-day registration for the 2016 gen*757eral election. In total, therefore, nearly 85% of the state’s population had access to this convenience, not counting those living in smaller counties whose regular polling place was the election authority’s main office or at the designated location in a city with more than 20% of the county’s population.
Plaintiff Patrick Harlan was, at the time this suit was filed, the Republican candidate for the United States House of Representatives for Illinois’s 17th Congressional District, which lies along the state’s western border. The Crawford County Republican Central Committee, which is also a plaintiff, is located in the 15th District, in southeastern Illinois. We refer .to them collectively as “Harlan,” as there is no material difference in the arguments they are advancing. On August 4, 2016—more than 18 months after P.A. 98-1171 became law, about four and a half months after the March 2016 primary (which was administered in accordance with P.A. 98-1171), and only 96 days before the general election— Harlan filed this action. He contended that P.A. 98-1171, codified at 10 ILCS 5/4-50, violates the Equal Protection Clause of the Fourteenth Amendment to the U.S. Constitution, because (as he sees it) the law disadvantages voters in smaller counties that do not have e-pollbooks, and thus comparatively boosts Democratic voter turnout.
On August 9, Harlan moved for a preliminary injunction barring election officials throughout the state from implementing the Election-Day registration system. The time for waiting apparently had expired; Harlan wanted the court to hurry. His motion argued that the law needed to be enjoined immediately so that no county would offer precinct-level registration in the upcoming November election, but he placed no time limit on the requested injunction. In support of his. request, he offered only the opinion of M.V. Hood, a political scientist at the University of Georgia. Hood reported that Illinois is the only state that uses this two-tiered system. He also said that research shows that Election-Day registration has a positive effect on turnout; that- the positive effect is not consistent across all age and residential groups; and that it is “quite possible” that Democratic candidates would be disproportionately benefited. Conspicuously absent from his report was data about eligible but not registered voters, their party affiliation, the distribution of those potential voters across large- and small- population counties, or some quantification or description of the burden (such as the distance) created by the tiered system. Harlan provided no other support—not even a single affidavit from a would-be voter.
The district court granted the'requested relief on September 27, 2016. The preliminary injunction ordered the election authorities in all 102 counties not to implement the Eléction-Day registration option at precinct polling places. It contained no language limiting it to the 2016 election. Defendants—the Chairman of the Illinois State Board of Elections, the other members of the Board, and the Cook County Clerk (who had intervened)—filed a notice of appeal that day, and also asked the district court to stay the preliminary injunction pursuant to Federal Rule of Civil Procedure 62(c). (We refer to the defendants collectively as the Election Officials.) The district court denied their motion on September 29; they renewed it in this court, and we granted the stay on October 4. We also denied expedited briefing and consolidated the appeals from the state officials (No. 16-3547) and the intervenor-defendant (No. 16-3597).
II
We heard oral argument in these appeals on May 30, 2017, more than five *758months after the 2016 general election. That naturally raises the question whether the case is moot: the election for the 17th Congressional District in which Harlan was running is over, the Democratic candidate (Cheri Bustos) won decisively, and the primaries for the 2018 congressional .races will not take place until next year. By that time, the counties encompassed within the 16th and 17th Districts may have eléctronic pollbooks, eliminating any concern that voters (Republican or otherwise) in those areas lack some registration and voting options. Harlan takes the position that the case was indeed rendered moot by the occurrence of the November 8, 2016, general election. The Election Officials point out that the preliminary injunction was hot set to end following that election, nor did the district court’s order self-destruct at that. point. Indeed, although the district court’s opinion referred to the then-upcoming election, it . also discussed the law’s ramifications on elections generally—a topic in which the Crawford County Republican Central Committee is undoubtedly interested. Although Harlan says that we should read a termination date into, the district court’s injunction, we are wary of doing so. Such an action would be inconsistent with Federal Rule of Civil Procedure 65(d), which requires all injunctions to be self-contained documents. We think it best to take the injunction at face value and on that basis determine what, if anything, is left of this case. '
In order to obtain a preliminary injunction,. a plaintiff must show three things: (1) without such relief, he will suffer irreparable harm before his claim is finally resolved; (2) he has no adequate remedy at law; and (3) he has some likelihood of success on the merits. Girl Scouts of Manitou Council, Inc. v. Girl Scouts of U.S. of Am., Inc., 549 F.3d 1079, 1086 (7th Cir. 2008). If the plaintiff can do that much, the court must then weigh the harm the plaintiff will suffer without an injunction against the harm the defendant will suffer with one. Ty, Inc. v. Jones Grp., Inc., 237 F.3d 891, 895 (7th Cir. 2001). In addition, the court must ask whether the preliminary injunction is in the public interest. Jones v. Markiewicz-Qualkinbush, 842 F.3d 1053, 1058 (7th Cir.-2016). This type of relief must not lightly be granted: the movant bears the burden of showing that it is warranted. Mazurek v. Armstrong, 520 U.S. 968, 972, 117 S.Ct. 1865, 138 L.Ed.2d 162 (1997). On appeal, however, we ask only whether the district court’s decision' to grant or deny a preliminary injunction represented an abuse of discretion. Lawson Prods., Inc. v. Avnet, Inc., 782 F.2d 1429, 1437 (7th Cir. 1986).
Taking the basic criteria in order, we look first at Harlan’s showing of harm. His case rests exclusively on Hood’s testimony, which we summarized earlier. We conclude that his opinion comes nowhere close to demonstrating that Illinois voters would suffer any harm at all—let alone irreparable harm—under the system set up by P.A. 98-1171. Hood’s report barely considered the effect pf the statute against' the backdrop of the rest of Illinois’s electoral system. At best, it gave a 30,000-foot view , of some academic research and a table of votes for the candidates of each major party in statewide races in small and large counties. The district court did not explain how that data related to the question before it.
Relying on a report from political scientists Roger Lorocca and John S. Kleman-ski, Hood admitted that political science research generally concludes that Election-Day registration has a positive effect on voter turnout. The same report indicated that “positive turnout effect ... is less consistent across age. and - residential groupings” when centralized Election-Day registration is used. Hood never said, however, which groups of voters—older, youn*759ger, rural, urban, ethnic, etc.—are statistically more or less likely to benefit from the same-day option. Hood also provided no information on the question whether Election-Day registration in Illinois’s 20 more heavily populated counties is more likely to increase voter participation than centralized Election-Day registration in the smaller counties. Hood also failed to take early voting into account, other than to say that it has been linked to negative turnout in studies from states other than Illinois. Lastly, Hood’s conclusion was far from definitive. He ended with these observations, which we quote verbatim:
• Limiting access in 82 of the state’s counties ... will likely dampen any positive turnout effect relative to larger counties where [Election-Day registration] will be implemented-in all voting precincts.
• It is quite possible ... that Illinois’s [Election-Day registration] scheme will have the added effect of diminishing GOP votes.
But Hood did not quantify his predictions. As a result, the court was in the dark about just how weak the “positive turnout” would be in small counties or how comparatively diminished Republican votes might be. And again, even though Hood purported to assess the impact of the voter registration law on Illinois counties, his report contained no facts about voter registration rates by county or municipality. That said, Hood did admit that the 20 counties required to provide Election-Day registration contain about 84 percent of the state’s population. His testimony falls far short of establishing a basis for finding that the law severely burdens the smaller-county residents. Without any other basis for finding irreparable injury—and there was none—the district court erred by deeming this element of the preliminary injunction test satisfied.
If, contrary to our conclusion, there had been some reason to find irreparable harm, we would move on to the question whether anything but a preliminary injunction could remedy that harm. We assume for the sake of argument that the answer to this question could be no—in other words, in that counterfactual world, immediate relief would have been needed. •
Last, we turn to likelihood of success on the merits. Here, too, we find nothing in this record that supports a finding that Harlan has any realistic chance of prevailing. We begin with some background rules. In a pair of cases, the Supreme Court addressed the constitutional rules that apply to state election regulations. See Anderson v. Celebrezze, 460 U.S. 780, 103 S.Ct. 1564, 75 L.Ed.2d 547 (1983); Burdick v. Takushi, 504 U.S. 428, 112 S.Ct. 2059, 119 L.Ed.2d 245 (1992); see also Crawford v. Marion Cnty. Election Bd., 553 U.S. 181, 128 S.Ct. 1610, 170 L.Ed.2d 574 (2008). Under the Anderson-Burdick test, the court must apply a flexible standard that depends on the severity of the burden imposed by the state law under consideration:
A court considering a challenge to a state election .law must weigh the character and magnitude of the asserted injury to the rights protected by the First and Fourteenth Amendments that the plaintiff seeks to vindicate against the precise interests put forward by the State as justifications for the burden imposed by its rule, taking-into consideration the extent to which those interests make it necessary to burden the plaintiffs rights.
Burdick, 504 U.S. at 434, 112 S.Ct. 2059 (quotation marks omitted). In Crawford, a plurality of the Court reiterated that “a court must identify and evaluate the interests put forward by the State as justifications for the burden imposed by its rule, *760and then make the ‘hard judgment’ that our adversary system demands.” 553 U.S. at 190, 128 S.Ct. 1610 (plurality opinion). (The other three Justices concurred in the judgment on the ground that the burden imposed by Indiana’s voter ID law imposed no special burden on voters and thus the state law was constitutional. See id. at 204, 128 S.Ct. 1610 (separate opinion of Scalia, J.).
The Court acknowledged that election.laws “inevitably affect[] ... the individual’s right to vote and his right to associate with others for political ends.” Burdick, 504 U.S. at 433-34, 112 S.Ct. 2059. When an election law severely burdens voters’ constitutional rights, that law must be narrowly drawn to advance a compelling state interest. Id. But when the law “imposes only reasonable, nondiscriminatory restrictions upon the rights of voters, the State’s important regulatory interests are generally sufficient to justify the restrictions.” Id. at 434, 112 S.Ct. 2059 (internal citation omitted). Regulations falling somewhere between those extremes—ie. those that impose more than a minimal, but less than a severe, burden on voting— are evaluated by comparing the burden on the voters with the state’s interest and the means it has chosen to advance that interest,
In our case, the district court applied strict scrutiny because it was persuaded that the burden on voters in the smaller counties was severe. But it is unclear on what, precisely, the court relied in arriving at that conclusion. For the reasons we already have reviewed, Hood’s contribution did not fill that void. Also troubling is how Harlan’s attack has shifted as this case has progressed. Hood’s report makes politics front and center; its only' analysis of P.A. 98-1171 comes in a section titled “The Partisan Effects of Illinois’ Election-Day Registration Statute.” This section discusses the vote totals garnered by the two major parties, by large and small counties. But on appeal Harlan concerns himself with the impact of the law on voters in low-population and rural counties, as well as less affluent counties, with only passing reference to partisan effects. But affluence is not a factor mentioned in the law, and nothing suggests that it is correlated with county size. We note that the state’s second-largest county is DuPage County, which abuts Cook County and Chicago. DuPage County includes some of the Midwest’s most affluent communities and has traditionally been a Republican stronghold. See DuPage County, Illinois, https://en.wikipedia.org/wiki/Du Page_County,-Illinois. This indicates that there is no necessary correlation between affluence, county size, and a tendency to vote Democratic. Finally, even if Harlan’s focus had remained steady, problems would remain. The question whether an election regulation places an unacceptable burden on voters is “not a factual finding, but a legal determination subject to de novo review.” Ohio Democratic Party v. Husted, 834 F.3d 620, 628 (6th Cir. 2016).
The Supreme Court held in both Burdick and Crawford that it is wrong to assume that any regulation of the voting process, no matter how trivial, must be assessed using strict scrutiny, just because voting is a fundamental right. In fact, Bur-dick squarely holds otherwise: “Petitioner proceeds from the erroneous assumption that a law that imposes any burden upon the right to vote must be subject to strict scrutiny. Our cases do not so hold.” 504 U.S. at 432, 112 S.Ct. 2059. It follows that not every difference in treatment amounts to an Equal Protection violation. See Crawford, 553 U.S. at 191, 128 S.Ct. 1610 (opinion of Stevens, J.).
Perhaps recognizing that it was on thin ice, the district court also concluded that even under rational-basis review, Har*761lan had some likelihood of succeeding on the merits. But it provided no support for this conclusion, and we cannot find any in the preliminary-injunction record. Harlan offers no evidence of discriminatory intent, as opposed to evidence of some differences in treatment. As the Supreme Court noted in Anderson, any state election law, “whether it governs the registration and qualifications of voters, the selection of eligibility of candidates, or the voting process itself, inevitably affects—at least to some degree—the individual’s right to vote and his right to associate with others for political ends.” 460 U.S. at 788, 103 S.Ct. 1564.
In light of the minimal harm posed by P.A. 98-1171 and Harlan’s failure to show a likelihood of success on the merits, we conclude that the preliminary injunction should not have issued.' Our view is bolstered by the absence of any consideration of the public interest in the opportunity to register to vote, and to vote. Even though P.A. 98-1171 does not force quite as many options on the smaller counties as it does on the 20 largest counties, it permits every county to adopt the default same-day rules, and it provides realistic same-day options even in the smaller places. This, coupled with the lack of any- data about which groups are disadvantaged and how, dooms the injunction.
Ill
Even if the preliminary injunction had been justified at the time the district court entered it, the passage of time has changed at least one thing: the urgency of injunctive relief. There is plenty of time at this juncture for the district court to consider this case without invoking its extraordinary equity powers. The next election that -will be affected by P.A. 98-1171 is the off-year primary election scheduled for March 20, 2018. To the extent that this injunction is designed to address Illinois elections until the merits of the case are resolved, we -conclude that it must be vacated.
The preliminary injunction entered by the district court on September 27, 2016, is hereby Vacated and the case is remanded to the district court for further proceedings consistent with this opinion.