This appeal challenges the decisions of two district courts holding that the practice by Illinois County Clerks of placing their own political party in the first or top position on voting ballots in all general elections violates the equal protection clause.
I
This appeal actually involves three different suits brought against three different County Clerks or County Boards of Election Commissioners in Illinois. Initially, there is the consolidated appeal of two district court cases, Sangmeister v. Woodard (Will County) and Culliton v. DuPage County Board of Election Commissioners (DuPage County), reported at 419 F.Supp. 126 (N.D.Ill.1976). However, the Sangmeister case was itself a consolidated trial of two actions; the other suit was Walsh v. Board of Election Commissioners (Cook County).
A
Although the cases were consolidated for this appeal, the Culliton case is in a somewhat different procedural posture before this court and so it will be discussed separately. Culliton was filed and tried prior to the November 1976 general election. Plaintiff was a Democratic Party candidate for State’s Attorney of DuPage County, Illinois. Defendant is a body of three Commissioners, the County Clerk as Chairman and two others selected by the Circuit Court of DuPage County — one person representing each major party — in accordance with Ill.Rev.Stat., 1975, ch. 46, § 6-21. Thus, the majority of that Board represents the Republican Party.
The plaintiff’s complaint alleged that the defendant Board “has always placed the Republican Party candidates on the first line of the ballot, and the Democratic Party candidates and the Independent candidates have always been placed in the second or on lower lines of the ballot.” The complaint characterized this practice as “arbitrary, preferential, discriminatory treatment” in violation of the equal protection clause of the Constitution and requested for relief, inter alia, a preliminary injunction enjoining the defendants from deciding the place*463ment of candidates on the ballot in the November 1976 election by unconstitutional means.
A hearing was held on September 16, 1976 which the Docket describes as a “hearing held on motion for preliminary injunction.” 1 Subsequent to the hearing the District Court entered on the Docket for September 20, 1976 “[pjreliminary injunction granted for plaintiff with opinion to follow.” In that opinion, the district court accepted the expert testimony of Dr. Samuel A. Kirkpatrick, professor of political science at Oklahoma University, to the effect that there was a favorable effect created by first or top ballot placement. The court also found that plaintiff had been intentionally denied equal access to the top ballot position. Based on those two findings, the district court held that the Board’s practice did violate the Constitution.
The district court’s remedy required the Board of Election Commissioners to employ an impartial, public drawing to decide ballot placement for each party. In addition, the court ordered the defendants to devise a rotational ballot placement procedure for future elections.
A public lottery was used by the DuPage County Board of Election Commissioners for placing candidates on the ballot in the November 1976 general election. The Board now appeals from the rest of the district court’s order.
B
Plaintiffs in Sangmeister were all candidates for office in Will and Cook Counties in the November 1976 general election. They all alleged that the practice of the Clerk in Will County and the Board in Cook County of placing the party of the Clerk in the first position on the ballot in general elections was unconstitutional and sought preliminary and permanent injunctions against the practice.
A hearing for a preliminary injunction was held on October 15, 1976 and the injunction was denied on the basis that there was too little time between the hearing date and the general election to award any relief.
On December 20, 1976, a trial was held before the district court. At that trial there were four witnesses: one expert for each side and the County Clerks of Will and Cook County. The plaintiffs’ expert, Dr. Samuel A. Kirkpatrick2 testified based on previous empirical studies done by other scholars3 and his own research based on the Will County May 1976 primary election, “that on the average first place garners 3.3 percent more votes than second place.” (Tr. at 110.) Dr. Kirkpatrick did recognize that there is a difference in ballot effect between primary and general elections and that he “would expect it to be somewhat reduced, perhaps by a percentage point, sometimes, maybe, as much as two percentage points.” (Tr. at 118.)
The defendants’ expert, Richard G. Smol-ka, a professor of government at American University, testified, based on his review of all previous studies on the effect of ballot placement on voting, that “there is virtually nothing at all been done on the subject much less anything been shown,” and thus *464“there is absolutely no evidence upon which to base an opinion.” (Tr. at 241.) He also testified that a rotational system for ballot placement would be a bigger administrative burden than it was worth given the lack of proven ballot placement effect on voting. (Tr. at 268.)4
Both of the County Clerks, Clara Hartley Woodard of Will County and Stanley T. Kusper, Jr. of Cook County were called by plaintiffs as adverse witnesses. Both witnesses testified that they had always placed their party at the top of the ballot for county offices in accordance with the customary procedure of the County Clerks in Illinois. (Tr. at 13-14, 26-28.) They both felt that such a practice was permissible under Ill.Rev.Stat, 1975, ch. 46, § 16-3 which states:
The list of candidates of the several parties and any such list of independent candidates shall be placed in separate columns on the ballot in such order as the authorities charged with the printing of the ballots [County Clerks] shall decide
At the end of the hearing, the district court orally rendered its decision in favor of the plaintiffs. The court held that “the position on the ballot for that candidate would be a preferred position on a general election even though he might appear under a party name.” (Tr. at 415.) In addition, the court found that the record “shows a discrimination against the candidates who are deprived of the top position on the ballot.” (Tr. at 415.)
To remedy the constitutional violation, the court required a
rotation of the major parties’ position on the general election ballot by precincts, even precincts with one major party at the top and the odd precincts with the other major party at the top.
(Tr. at 420).5
The defendants have appealed both the court’s finding of a constitutional violation and its remedy requiring the County Clerks to adopt a rotational system for deciding future ballot placement.
II
In the Culliton case there is some dispute regarding the exact nature of the district court’s order. The hearing was set on a motion for a preliminary injunction and the Docket reads that a preliminary injunction hearing was held and that the court granted plaintiff’s motion for a preliminary injunction. Nonetheless, the district court’s memorandum opinion could be read as prescribing more permanent relief to plaintiff.6 This court declines to so read the district court’s opinion. It is our view that this case involves merely an appeal of the lower court’s grant of a preliminary injunction as provided in 28 U.S.C. § 1292(a)(1) (1970). We affirm the preliminary injunction against defendants and remand to the district court for further proceedings on a permanent injunction to be conducted in accordance with the views expressed in the remaining portions of this opinion.
In reviewing the district court’s order for a preliminary injunction, this court *465is mindful that it has “very limited scope of review,” Gillespie & Co. v. Weyerhauser Co., 533 F.2d 51, 53 (2d Cir. 1976) (per curiam), and must merely “satisfy ourselves that the findings of the trial judge are not clearly erroneous, and that he has not abused the broad discretion which is his in determining whether to grant or withhold interlocutory relief.” Brandeis Mach. & Supply Corp. v. Barker-Greene Co., 503 F.2d 503, 505 (6th Cir. 1974) (per curiam). In addition, since this relief is merely interlocutory and thus this case is yet to be heard in full on the merits, this court should “refrain from unnecessary comment on the evidence.” Id.
Examining the record in this case, it seems clear that plaintiff has “raised questions going to the merits so serious, substantial, difficult and doubtful, as to make them a fair ground for litigation . . .” Hamilton Watch Co. v. Benrus Watch Co., 206 F.2d 738, 740 (2d Cir. 1953); Brandeis Machinery, supra at 505. Plaintiff has brought in testimony that tends to establish that ballot position does have an effect on voting. Also, the testimony of the defendant County Clerk could be interpreted to suggest a discriminatory practice exists. Thus, the evidence supports fully the district court’s preliminary injunction. A full hearing on the merits should now be held in accordance with the views expressed in the remainder of this opinion.
Ill
It has become well established that the power of the state to regulate elections must be exercised consistent with the dictates of the equal protection clause of the Fourteenth Amendment. Williams v. Rhodes, 393 U.S. 23, 89 S.Ct. 5, 21 L.Ed.2d 24 (1968); Communist Party v. State Bd. of Elections, 518 F.2d 517 (7th Cir. 1975); Anderson v. Schneider, 67 Ill.2d 165, 365 N.E.2d 900 (1977). In determining the constitutionality under the equal protection clause of ballot placement procedures, this court in Bohus v. Board of Election Commissioners, 447 F.2d 821 (7th Cir. 1971), adopted a two-part test. First, the plaintiff must “show that top placement on the ballot is an advantage in an election . .” Id. at 822. See also Weisberg v. Powell, 417 F.2d 388, 392 (7th Cir. 1969) (per curiam). Second, the plaintiff must “prove the existence of an intentional or purposeful discrimination by authorities in which one class is favored over another.” 447 F.2d at 822. See also Snowden v. Hughes, 321 U.S. 1, 8, 64 S.Ct. 397, 88 L.Ed. 497 (1944).
The effect of ballot placement on voting is a question of fact, and our review is limited to whether the district court’s finding was clearly erroneous within the meaning of Fed.R.Civ.P. 52(a). 447 F.2d at 823. In the Sangmeister case, we hold that the trial court’s conclusion that “top placement on the ballot would be an advantage to the plaintiff” is supported by substantial evidence and therefore is not clearly erroneous.
There were several studies presented to the district court by Dr. Kirkpatrick on the effect of ballot placement that reached the conclusion that a candidate placed first on a ballot acquires a substantial vote getting advantage. Dr. Kirkpatrick classified them into three general types of studies: field studies,7 laboratory studies8 and ex post facto research. Of these studies the most numerous and probative are ex post facto *466studies in which actual election results — in elections with rotated ballots — are examined. Dr. Kirkpatrick concluded that from all of the ex post facto studies a favorable first place position effect was found in between 75 and 100 percent of the elections studied. (Tr. at 130.) In Dr. Kirkpatrick’s study of the Will County 1976 Primary, he found that there was a positive first position ballot effect in eight of ten races. (Tr. at 111.) Almost all of these studies in their final conclusions support the district court’s finding.9
Defendants attempt to discount the effect of these studies by arguing that none of them deals with an Illinois general election. It is true that most of these studies deal with primary and multi-candidate elections, and thus there are differences between the elections studied and the Illinois general election. Nonetheless, merely because the studies are not perfectly suited to the facts does not dissipate completely their probative value.10 Moreover, defendants recognize that some of the studies deal with general elections and in those the general rule is that there is a first place position effect on vote getting.11 Finally, there are no unimpeached empirical studies that prove that there is no ballot placement effect.12 In fact, even defendants’ expert witness, Professor Smolka, did not so testify. He merely concluded as a social scientist that there was insufficient evidence upon which to draw a conclusion. (Tr. at 241.) We therefore conclude that there was no error in the district court’s finding that first position on a ballot is an advantage to a political candidate.13
*467There is some dispute about what is required by the second portion of the Bohus test, i. e., the need to show intentional discrimination. Bohus itself is not very helpful in determining the meaning of that test since the court did no more than state it.14
This court, however, did find an intentional discrimination in Weisberg v. Powell, 417 F.2d 388 (7th Cir. 1969) (per curiam). That case involved an equal protection challenge to the practice of Secretary of State Powell of selecting arbitrarily for the top ballot positions candidates he knew and felt would make good candidates for Constitutional Convention Delegates. Mr. Powell testified at the trial that he deliberately placed candidates on the ballot with the hope that those placed at the top of the list would be elected. Id. at 391 n. 4. Defendants argue that the intent requirement will only be satisfied by the types of statements made by Mr. Powell in the Weisberg case. They claim that since there is no testimony in this record expressly stating that the Clerk’s practice was designed to aid the Clerk’s political party, there can be no equal protection violation.
We disagree. Whatever the contours of the intent requirement are in general in equal protection analysis see Washington v. Davis, 426 U.S. 229, 240-42, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976), we believe that the facts in this case satisfy that requirement. We note initially that the practice at issue here has been carried out for at least 100 years in DuPage, 40 years in Will and 30 years in Cook County. We also note that it appears that every County Clerk in Illinois engages in the same practice, regardless of whether he or she is a member of the majority or minority party in that county.
Such systematic and widespread exclusionary practices by the State officials satisfy even the strictest intent requirement. For instance, in Snowden v. Hughes, 321 U.S. 1, 9, 64 S.Ct. 397, 401, 88 L.Ed. 497 (1944), the Court in explaining the intent requirement cited as proof of intent to tax unequally “a systematic under-valuation of the property of some taxpayers and a systematic over-valuation of the property of others.” Similarly, in Washington v. Davis, 426 U.S. 229, 242, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976), the Court considered what proof would be necessary to prove exclusionary intent in jury selection cases. The Court reasoned that “the total or seriously disproportionate exclusion of Negroes from jury venires — may for all practical purposes demonstrate unconstitutionality . . . .” The widespread and persistent practice of Illinois County Clerks to exclude opposition party members from top ballet positions is sufficient to satisfy the standards adopted by the Court in Snowden and Davis.
Defendants respond that this discrimination can be justified on the basis that it satisfies the administrative need to avoid confusion and to have a consistent practice so that voters will know in advance where the parties will be on the ballot. It is difficult to understand how this practice satisfies those requirements any more efficiently than would a neutral system of ballot placement. In light of this fact, we find that the County Clerks’ practice does not further any substantial state interest.
We believe that the evidence clearly supports the district court’s conclusion that the discrimination is intentional or purposeful. We, therefore, hold that the practice of Illinois County Clerks of excluding plaintiffs from top ballot positions was intentional and worked a substantial disadvantage to them in violation of the Fourteenth Amendment to the United States Constitution.
IV
Defendants contend that the district court abused its discretion in ordering the County Clerks to adopt a rotational system for determining ballot placement. *468We agree. It is our view that although the district court has broad powers to fashion an appropriate remedy in the public interest, Hecht Co. v. Bowles, 321 U.S. 321, 329-30, 64 S.Ct. 587, 88 L.Ed. 754 (1944); Brown v. Board of Education, 349 U.S. 294, 300, 75 S.Ct. 753, 99 L.Ed. 1083 (1955); Swann v. Charlotte-Mecklenburg Bd. of Education, 402 U.S. 1, 15, 91 S.Ct. 1267, 28 L.Ed.2d 554 (1971), the remedy in the Sang-meister case is unnecessarily restrictive.
There are several factors which mitigate against imposing a specific rotational system for ballot placement on the County Clerks. First, we are reviewing the administrative implementation of a state statute that delegated broad authority to these clerks. Presumably, the legislature had in mind that each clerk would adopt a ballot placement procedure well suited to the specific needs of each county. Since a rigid requirement that all clerks adopt a particular procedure might well undermine the State legislature’s intent, it would be a proper application of the principles of federalism to abjure such a remedy.
Second, there is no reason to believe that the County Clerks would fail to adopt constitutionally permissible means of ballot placement. As a general rule, courts of equity need not impose specific requirements absent some reason to believe that a less restrictive approach will fail to remedy the constitutional violation. Plaintiffs cite cases upholding busing to demonstrate how broad the remedial powers of a federal court are. Those cases, however, also demonstrate the need for evidence that state officials will continue to violate the constitution before specific remedies should be imposed. As the Court in Swann stated: “Judicial authority enters only when local authority defaults.” 402 U.S. at 16, 91 S.Ct. at 1276.15
Third, the district court, itself, was less than confident in its remedy. The court stated after the trial in its decision:
I would not even discourage a re-hearing particularly on the question of the actual mechanics or practicability as to how this decision can be best effectuated or implemented. We really haven’t had very much testimony on that.
(Tr. at 421.) In light of the fact that the issue is complex — for example, concern must be given to such problems as computer programming, type of voting device, possible combinations of ballots for printing, size of voting units, etc., — it is not clear how well a court could devise a ballot selection method even after an exhaustive hearing. This is an area where the County Clerks must be presumed to be expert, or at least to have expert assistance available on these various problems and so are best suited for developing a constitutionally valid and practicable ballot placement procedure.
We thus vacate the district court’s order that a rotational selection procedure be implemented by the County Clerks and remand the case to the district court to order the defendants to devise a constitutionally permissible ballot placement procedure. In adopting a procedure, the County Clerks should abide by the following guidelines:
1) The procedure adopted must be neutral in character. This court will not accept a procedure that invariably awards the first position on the ballot to the County Clerk’s party, the incumbent’s party (see Mann v. Powell, 333 F.Supp. 1261, 1267 (N.D.Ill.1969) (three judge court), or the “majority” party.
2) The procedure adopted should take account of all political parties involved, major and/or minor. While we need not decide whether the district court’s order to exclude minor parties from the rotational selection procedure was valid, we recognize that there may be serious questions about its constitutional propriety.
3) The County Clerks should feel free to adopt any constitutional procedure and *469to experiment from election to election if they feel such an approach is desirable.16
To summarize, we affirm the district court’s preliminary injunction in Gulliton and remand the case for further proceedings on the permanent injunction consistent with this opinion. We affirm the district court’s finding in Sangmeister that there has been a constitutional violation. We vacate the district court’s remedy in Sang-meister and remand the case to the district court to order the defendants to adopt a constitutionally permissible ballot placement procedure.
. When the Court Clerk announced the case at the hearing, he described it as a “motion for preliminary injunction.” (Tr. at 3.)
. This is the same Dr. Kirkpatrick who testified at the Culliton hearing for a preliminary injunction.
. See Kamin, Ethnic and Party Affiliations of Candidates as Determinants of Voting, 12 Canadian J.Psych. 205 (1969); Coombs, Peters & Strom, Bandwagon, Ballot Position and Party Effects: An Experiment in Voting Choice, 3 Experimental Stud.Pol. 31 (1974); Taebel, The Effect of Ballot Position on Electoral Success, 19 Am.J.Pol.Sci. 519 (1975); Brooks, Voters’ Vagaries: The Value of Position on a Ballot, 10 Nat’l.Municipal Rev. 161 (1921); White, Voters Plump for First on List, 39 Nat’l.Municipal Rev. 110 (1950); H. Bain & D. Hecock, Ballot Position and Voter’s Choice (1957); Mueller, Voting on the Propositions: Ballot Patterns and Historical Trends in California, 68 Am.Pol.Sci.Rev. 1197 (1969); W. Durley, Secretary of State’s Unpublished Statistical Table of 1970 and 1974 Primary Elections (1975); Note, California Ballot Position Statutes: An Unconstitutional Advantage to Incumbents, 45 So.Cal.L.Rev. 365 (1972).
. This was a conclusion that had been drawn earlier by a federal task force on which Professor Smolka had worked. See Office of Federal Elections, Study of Election Difficulties in Representative American Jurisdictions (1973).
. The district court restricted its remedy to major parties based on its view that it would be “unpractical and unrealistic” to include minor parties in a rotational placement system (Tr. at 418). The court reasoned:
We have to be practical in any attempt to balance the interests of various parties in litigation. When we are balancing them against the interest of the community at large and the practicality of an intelligible and meaningful election, I believe the minor parties must stand aside.
Id. One minor party — the Libertarian party— was represented at the trial, but did not appeal this portion of the district court’s order.
. The defendant claims that the district court merged a hearing on a preliminary injunction into one for a permanent injunction without sufficient notice in violation of Fed.R.Civ.P. 65(a)(2). See Pughsley v. 3750 Lake Shore Drive Cooperative Bldg., 463 F.2d 1055, 1057 (7th Cir. 1972). As shown below, we disagree with the defendant’s interpretation of the proceedings before the district court.
. In these studies the researcher goes into an electoral district just prior to an actual election and asks prospective voters to vote on a fictitious set of candidates. The only study of this kind cited by Dr. Kirkpatrick was Kamin, supra note 3. While the author of the study concluded that ballot placement did make some difference to those voters studied, its probative value is slight because it is over 20 years old, dealt with a Canadian Parliamentary election and was not really conducted to determine ballot placement effect.
. There have been two laboratory or voting simulation studies on ballot placement effect. See Coombs, et al., supra note 3; Taebel, supra note 3. Both of these studies found that there was a ballot placement effect on voting. The simulated vote attributable to ballot placement in one study was 5 percent (Coombs, et al.) and in the other, 6.3 percent (Taebel).
. See, e. g., Mueller, supra note 3, at 1208 (“the candidate who is listed first in low visibility races has a definite advantage”); Note, supra note 3, at 376 (“one can attribute at least a five percent increase in the first listed candidate’s vote total to a positional bias.”); H. Bain & D. Hecock, supra note 3, at 14 (“even in a general election, where the voters have the more clear-cut guidance provided by party labels in making their choice, there is a considerable advantage to be gained by appearing in certain positions.”).
. Such a restrictive view would almost totally eliminate any judicial use of the social sciences. Also, the Supreme Court has never taken so limited a position on the evidentiary uses of social science studies. In Brown v. Board of Education, 347 U.S. 483, 74 S.Ct. 686, 98 L.Ed. 873 (1954), the Court relied on several general social science studies in concluding that segregation was harmful to black children. Id. at 494 n. 11, 74 S.Ct. 686. Those studies were cited and relied on notwithstanding the fact that Professor Kenneth Clark had conducted specific doll tests on the black children who were the plaintiffs in one of the consolidated cases and testified as to the results of those tests. See R. Kluger, Simple Justice 315-345 (1976).
. Two of the studies are from the Note in the Southern California Law Review in which there was a 5.3% and a 6.1% first ballot position effect. Bain & Hecock present data on two general elections — a congressional election in Ohio and an election in Colorado — where they found generally that there was a first place ballot effect. See the conclusion they drew from their data quoted in note 9 supra.
. There were two studies mentioned at the trial that seemed to reach the conclusion that there was no first position effect on voting. Bagley, Does Candidate’s Position on the Ballot Paper Inñuence Voter Choice; A Study of the 1959 and 1964 British General Elections, 19 Parl.Affairs 162 (1966); Byrne & Pueschel, But Who Should I Vote for for County Coroner?, 36 J.Pol. 778 (1974).
In the Bagley study, there were two elections studied. In the initial data analysis, it was found that there was a ballot placement effect in one election but not in the other. According to Dr. Kirkpatrick, a subsequent data analysis, however, proved that there was a ballot position effect in both elections. (Tr. at 205-06). Thus, this study actually favors plaintiffs’ position.
The Byrnes and Pueschel study also concluded that there was no ballot position effect on voting. Dr. Kirkpatrick testified, however, that there was a serious methodological flaw in their research design. (Tr. at 207-08). It was his view that the study was deceptive because the researchers had chosen to study an aggregation of elections and not to examine each race separately. It was Dr. Kirkpatrick’s opinion that such an approach might well mask the existence of a first position effect in specific races. (Tr. at 208).
. We should note that this court is not the first to find such an effect. See, e. g., Kauten-burger v. Jackson, 85 Ariz. 128, 333 P.2d 293 (1958); Gould v. Grubb, 14 Cal.3d 661, 122 Cal.Rptr. 377, 536 P.2d 1337 (1975); Elliott v. Sec’y of State, 295 Mich. 245, 294 N.W. 171 (1940); Holtzman v. Power, 62 Misc.2d 1020, 313 N.Y.S.2d 904 (1970); State, ex rel. Roof v. Board of Commissioners, 39 Ohio St.2d 130, 314 N.E.2d 172 (1974).
. The court in Bohus affirmed the district court’s determination that there was no first place ballot effect as being not clearly erroneous. 447 F.2d at 823. Thus, the court never reached the intent issue.
. In evaluating the likely response of the County Clerks to this decision, we can draw no inference from the fact that we now find they have been violating the Constitution because this court approved this very practice less than seven years ago in Bohus.
. We note that in one of the exhibits at the trial there is included a list of methods by which political parties are assigned ballot positions in both primary and general elections throughout the fifty states. (Def. Ex. 9). One can only be struck by the diversity of approaches that have been taken by the various states. The County Clerks are encouraged to consider all of them in their search for a ballot selection procedure that is most appropriate for their respective counties.