788 P.2d 81
Ariz.
1990
NOEL FIDEL, Vice Chief Judge,

Court of Appeals.

In a contested primary election decided by an eight vote margin, sixteen illegal votes were cast. The loser claims in this petition for special action that the election must be set aside because one cannot know which of the candidates received the highest number of legal votes. We have taken jurisdiction to reexamine the law that governs elections when illegal votes exceed the margin of victory.

FACTS

In the 1988 primary election for Navajo County Attorney, petitioner Bret H. Huggins narrowly lost the Democratic Party nomination to real party in interest Dale K. Patton. After conducting a recount pursuant to A.R.S. § 16-661, the Secretary of State reported that Patton had won by 3,593 votes to Huggins’s 3,585. This eight-vote margin, however, was exceeded by sixteen votes illegally cast. Fifteen voters registered as independents or non-partisans had been improperly permitted to vote Democratic Party ballots. The sixteenth illegal voter was a convicted felon whose electoral rights were unrestored.

Huggins contested the election, as A.R.S. § 16-671 permits, but lost because he was unable to prove for whom the illegal votes were cast. Though he proved that illegal votes were cast in sufficient number to change the election result, he could not prove that they changed the result in fact. This special action arises from the trial court’s rejection of Huggins’s election challenge.

By order at the time this matter was submitted, we accepted jurisdiction but denied relief. We explain our ruling in this opinion.

The morgan-millet rule

A challenger’s burden of proving how illegal votes were cast derives in Arizona from Morgan v. Board of Supervisors, 67 Ariz. 133, 192 P.2d 236 (1948), where this court quoted the following statement from C.J.S.:

Legality of votes. Where an election is contested on the ground of illegal voting, the contestant has the burden of showing [1] that sufficient illegal votes were cast to change the result, and [2] of showing for whom or for what they were cast.

67 Ariz. at 143, 192 P.2d at 243 (quoting 29 C.J.S., Elections, § 274) (emphasis added).

Only the first of the quoted burdens was material in Morgan. Because the Morgan challengers failed to prove sufficient illegal votes to change the election result, the court had no need to determine what, if any, further burden might have faced them. The Morgan dictum, however, became holding in Millet v. Board of Supervisors, 6 Ariz.App. 16, 429 P.2d 508 (1967), where the court of appeals rejected an election challenge by a contestant who, like *350 Huggins, had carried the first burden, but not the second. 1

In this case, the trial court criticized, but felt obliged to follow, the Morgan-Millet rule. Huggins now urges us to abandon that rule and to relieve election contestants of the burden of proving how illegal votes were cast. Huggins directs us to Baggett v. State Election Board, 501 P.2d 817 (Okla.1972), where, under circumstances similar to these, the Oklahoma Supreme Court nullified an election, stating:

If election officials have not conducted an election according to law and knowingly permit non-registered Democrats to vote in a Democratic runoff primary election, the inexcusable conduct of the election officials should not inure to the benefit of any candidate either directly or indirectly.

Id. at 824. The trial court described Bag-gett as having “the force of reason behind it,” and Huggins urges us to make the Oklahoma approach our own.

We too see much reason in Baggett. Like the majority in that case, we recognize the inequity of burdening the challenger “to prove for which candidate the unlawful ballots were cast [in order] to be relieved from having the illegal ballots counted as legal ballots.” Id. Moreover, the challenger’s burden increases with the size of the unlawful vote. As Huggins argues persuasively, “it hardly seems fair that as the amount of illegal voting escalates, the likelihood of redressing the wrong diminishes.”

There are additional difficulties with the Morgan-Millet rule, which stem from the need to prove how illegal votes were cast through the testimony of those who cast them. First, as Justice Jackson pointed out in concurrence in Baggett, voters who have cast unlawful ballots may choose to assert their fifth amendment privilege not to testify. Id. at 825. The resulting exacerbation of the challenger’s burden is even more pernicious than Justice Jackson described. Though an illegal voter might be motivated to maintain silence by a genuine fear of criminal sanctions, a supporter of the challenger’s opponent might equally be motivated by the recognition that an invalid vote against the challenger would likely be cancelled only if the voter revealed how it was cast. Thus, the Morgan-Millet rule not only burdens a challenger onerously; it actually empowers partisans of the opposition to frustrate an election challenge and preserve illegal votes by exercising fifth amendment rights.

There is a second and related weakness to the Morgan-Millet rule. Voter disclosure testimony, even where offered, is highly suspect. Courts have long recognized this weakness when contemplating testimony by legal voters whose attempted votes were erroneously unrecorded. As the Utah Supreme Court stated:

We know from common experience that those who do vote are usually unwilling that the character of their votes be made public, and that whenever there is an investigation as to the actual vote cast it is almost certain to bring about prevarication and uncertainty as to what the truth is____ The temptation to actual fraud and corruption on the part of the candidates and their political supporters is never so great as when it is known precisely how many votes it will take to change the result____

Young v. Deming, 9 Utah 204, 212, 33 P. 818, 820-21 (1893); see also Briscoe v. Between Consol. School Dist., 171 Ga. 820, 824, 156 S.E. 654, 656 (1931) (“[I]t would ... be dangerous to receive and rely upon the subsequent statement of the voters as to their intentions, after it is ascertained precisely what effect their votes would have upon the result.”); Babnew v. Linne-man, 154 Ariz. 90, 93-94, 740 P.2d 511, 514-15 (App.1987) (same).

We concur in these comments and attribute comparable weakness to the testimony of illegal voters asked to disclose accomplished votes. See McCavitt v. Reg *351 istrars of Voters, 385 Mass. 833, 849, 434 N.E.2d 620, 630-31 (1982).

There is a third and especially troublesome problem associated with the Morgan-Millet rule: the prospect of judges compelling good faith voters who have cast invalid ballots to reveal what they supposed were private votes. Division Two of the court of appeals has recently approved such compulsion in Babnew, 154 Ariz. at 95, 740 P.2d at 516 (“If a witness discloses that he did vote, and it is shown clearly to be an illegal vote, then he may be compelled to disclose for whom he voted.”). The McCavitt opinion records the outrage of a voter so compelled: “I will not answer that question because, as far as I’m concerned, that is illegal. Nobody has the right to know who I voted for.” 385 Mass, at 846 n. 16, 434 N.E.2d at 629 n. 16.

Massachusetts rejects compelling voter testimony in these circumstances as “a kind of inquisitorial power unknown to the principles of our government and constitution.” Id. 434 N.E.2d at 630 (quoting Johnston v. Corp. of Charleston, 1 S.C.L. (1 Bay) 441 (1795)). The court continues: “We cannot sanction a result which tends to reduce citizen participation in the election process. That is too high a price to pay in a participatory democracy.” McCavitt, 385 Mass. at 848, 434 N.E.2d at 630.

This criticism strikes a responsive chord in Arizona, where our constitution explicitly assures secrecy in voting. 2 We need not now determine whether, under any circumstances, our constitutional commitment to ballot secrecy might accommodate compelling good faith voters to disclose invalid votes. It is sufficient for present purposes to recognize the force of that commitment and to explore alternative solutions that permit us to avoid compulsion so offensive to democratic sensibilities and assumptions. See Clay v. Town of Gilbert, 160 Ariz. 335, 343, 773 P.2d 233, 241 (App.1989) (Fidel, J., concurring) (commending avoidance of “the democratic anathema of compelling good faith voters to reveal their votes”).

NULLIFICATION AND RESUBMISSION

The solution commended by Huggins is to nullify the contested election and to order a new election when, as here, a challenger has proven that the margin of victory is exceeded by the number of invalid votes. This solution, the one chosen in Baggett and McCavitt, permits the public a second effort to achieve a properly conducted election.

A second election, however, is not immune from illegal ballots and may prove no better than the first. Moreover, a second election is costly, and the costs are not limited to the heavy fiscal expense of running an election another time. Some votes will be lost in a second election that were properly recorded in the first; these include voters who have died, voters who have moved, and voters whose interest in the office or electoral issue is too attenuated to pull them to the polls a second time. Additionally,

there may ... be identifiable biases in second elections. Candidates with ready access to financing and with strong and continuing party organizations will be able to mobilize a second campaign in the short time available much more effectively than opponents who lack such advantages. Candidates with support concentrated among less active voters may be disadvantaged in a second election if such supporters do not turn out to cast ballots when only one office is at stake.

Note, Developments in the Law: Elections, 88 Harv.L.Rev. 1111, 1315 (1975).

These costs and biases make us hesitant to nullify first elections automatically upon proof that the winner’s margin of victory was exceeded by the number of illegal votes.

*352 THE GROUNDS RULE

Oddly, we find a better alternative in a case that this court issued in the same year as Morgan—Grounds v. Lawe, 67 Ariz. 176, 193 P.2d 447 (1948). In Grounds, unlike Morgan, illegal votes were cast in sufficient number to change the outcome; thus, the issue was squarely presented whether to require the challenger to prove how they were cast. We de? dined to place this burden on the challenger, but also declined to set the election aside. Grounds, 67 Ariz. at 184-85, 193 P.2d at 452-53. Instead, we considered whether the outcome would be altered by a proportionate, precinct-by-precinct extraction of the illegal votes. That is, for each district in which invalid votes were cast, we calculated a “pro rata deduction of the illegal votes according to the number of votes cast for the respective candidates in [that] election district.” Id. at 182, 193 P.2d at 451. 3 Because the illegal votes were insufficient, when extracted in this fashion, to change the election result, the victory of the declared winner was confirmed.

Grounds, though less often cited than our Morgan dictum, was recently rediscovered by the court of appeals in Clay v. Town of Gilbert, 160 Ariz. at 338-39, 773 P.2d at 236-37. Grounds has also been cited with approval by the Supreme Court of Alaska, Hammond v. Hickel, 588 P.2d 256, 260 (Alaska 1978), and we reaffirm its application in this case.

We recognize an arbitrary element to proration. As we said in Grounds: “[T]he truth might be, if it could be shown, that all the illegal votes were on one side, while it is scarcely to be presumed that they would ever be divided between the candidates in exact proportion to their whole vote.” 67 Ariz. at 183, 193 P.2d at 451-52, quoting McCrary on Elections, 4th Ed., §§ 495-97. This element has led the Supreme Judicial Court of Massachusetts to reject proration in favor of nullification and resubmission to the voters “ ‘whenever the irregularity ... of the election is such that the result ... would be placed in doubt.’ ” McCavitt, 385 Mass, at 850, 434 N.E.2d at 631 (quoting Callison v. Peeples, 102 S.C. 256, 265, 86 S.E. 635, 637 (1915)).

The Supreme Court of Alaska, however, has recognized that proration is a useful “analytical tool ... for the limited purpose of determining whether ... [there] were errors of sufficient magnitude to change the result____” Fischer v. Stout, 741 P.2d 217, 226 (Alaska 1987). Alaska has not yet determined what it will do if a case arises where “the election result is put in doubt by application of the proportionate reduction rule.” Id. at 226 n. 15. Where, however, as in Grounds and Clay, proportionate reduction does not change the result, Alaska certifies the declared winner of the first election. Id. at 226.

Huggins challenges the Grounds approach as constitutionally invalid. Article VII, § 7 of the Arizona Constitution, the source of Huggins’s argument, provides: “in all elections held by the people in this state, the person, or persons, receiving the highest number of legal votes shall be declared elected.” Huggins reasons from this provision that, because proration does not permit us to be certain who won the highest number of legal votes, it is constitutionally proscribed.

We disagree. Justice Holmes cautioned long ago against pressing constitutional provisions to their “grammatical extreme.” “[S]ome play must be allowed to the joints if the machine is to work.” Tyson and Brother v. Banton, 273 U.S. 418, 446, 47 S.Ct. 426, 434, 71 L.Ed. 718 (1927) (Holmes, J., dissenting). See also Missouri, Kansas *353 & Texas Ry. Co. v. May, 194 U.S. 267, 270, 24 S.Ct. 638, 639, 48 L.Ed. 971 (1904) (“Great constitutional provisions must be administered with caution. Some play must be allowed for the joints of the machine....”).

The problem we confront is practical; the solution we choose is “workable.” Clay, 160 Ariz. at 338, 773 P.2d at 236. The Arizona Constitution, in our view, permits us room to make this choice. While proration is imperfect, we lack the luxury of perfection, and proration strikes us as a sensible screening device in a multi-district case. 4 First, proration spares the body politic the offensive voter compulsion of the Morgan-Millet rule. Second, when limited as in Fischer, it permits us at least sometimes to avoid the cost and delay of a second election.

Moreover, though proration leaves some doubt that we have discovered the true winner, the other options fail to bring us nearer to that mark. The practical impact of the Morgan-Millet rule, with its virtually impossible burden on the challenger, is to let illegal votes count. The nullification remedy invalidates a multitude of first election legal votes, passes the choice to the inevitably different electorate that turns out for a second election, and accepts the second election biases and distortions earlier described. Proration, by comparison, has the virtue of neutrality; and in election contests, neutrality is a major constituent of fairness.

Like the Supreme Court of Alaska, we defer deciding what must be done when proration would change an election result. The law advances incrementally; we address the increment before us; the legislature may wish to consider the subject before we visit it again. For now, we reaffirm the Grounds approach as a limited screening device. When, as here, the margin of electoral victory is exceeded by the number of invalid votes and the invalid votes were cast in more than one precinct, the impact of those votes shall be tested by proportionate deduction. When, as here, proportionate reduction does not change the result, the declared victory may be confirmed.

CONCLUSION: PRORATION APPLIED

Proration in this case changes the tally in eight Navajo County precincts, taking six votes from Patton’s overall tally and nine from that of Huggins. 5 By this method *354 Patton remains the winner with 3,587 votes to 3,576 for Huggins. Thus, we conclude that Huggins’s election challenge was properly denied.

FELDMAN, V.C.J., and MOELLER and CAMERON, JJ., concur. GORDON, C.J., did not participate in this decision; pursuant to Ariz.Const. aH. 6, § 3, NOEL FIDEL, Vice Chief Judge, Court of Appeals, Division One, was designated to sit in his stead. WILLIAM A. HOLOHAN, J., participated in this matter but retired prior to the filing of the opinion. CORCORAN, J., did not participate in the determination of this matter.

Notes

1

. In Moore v. City of Page, 148 Ariz. 151, 713 P.2d 813 (App.1986), the court of appeals reiterated the Morgan-Millet rule. In Moore, however, as in Morgan, the challengers’ failure to meet the first burden made it unnecessary to consider the suitability or fairness of the second.

2

. Ariz. Const, art. VII, § 1 provides: "All elections by the people shall be by ballot, or by such other method as may be prescribed by law; provided, that secrecy in voting shall be preserved.” See also Ariz. Const, art. II, § 8, which provides: "No person shall be disturbed in his private affairs, or his home invaded, without authority of law.”

3

. For example, consider a pair of hypothetical precincts with a mathematically convenient turnout of 100 voters in each. In the first precinct, Smith receives 60 votes, Jones receives 40, but 10 invalid votes are cast. Pro rata deduction would reduce Smith’s tally by 6 votes (60% of the invalid votes) and Jones’s tally by 4 (40%). In the second precinct, Jones wins 80 votes. Smith wins 20, but 5 invalid votes are cast. Here, pro rata deduction takes 4 of the invalid votes (80%) from Jones and 1 (20%) from Smith. Application of this method precinct-by-precinct is a neutral method to extract invalid ballots from the overall total. As we point out infra, however, it only functions neutrally in elections where more than one precinct has experienced invalid votes. See footnote 4.

4

. We observed in Grounds, and the court of appeals reiterated in Clay, that proration can only work fairly where more than one district has undergone invalid ballots. To make a prorated deduction of invalid ballots in a single district election would functionally amount to ignoring the invalid votes. Grounds, 67 Ariz. at 184, 193 P.2d at 452; Clay, 160 Ariz. at 339 n. 2, 773 P.2d at 237 n. 2. Thus, we only announce a rule today for multi-district elections. We will deal with single district elections when the need arises.

5

. Of the sixteen illegal votes, we can account for the precincts of fifteen. As for the sixteenth, which was cast by absentee ballot, our limited special action record does not tell us the precinct from which it was cast. This is a record limitation, not a systemic one, however, as the precinct of an absentee voter is statutorily required to be recorded. A.R.S. § 16-542(D).

Our inability to identify the precinct of one illegal voter is fortunately inconsequential in this case. Proportionate deduction of the fifteen votes identifiable by precinct establishes Patton’s victory by a margin of eleven.

The deductions are calculated as follows:

Illegal Votes Deducted from Patton's Total
Precinct Illegal Votes Percent of Vote Rec'd Deduction
Bird Springs 2 31.6 .632
Black Mesa 1 65.6 .656
Chilchinbeto 2 48.7 .974
Cibecue 3 22.8 .684
Low Mountain 1 47.0 .470
Pinon 2 61.1 1.222
Whitecone 2 48.4 .968
White River 2 36.7 .734
VOTES DEDUCTED FROM PATTON’S TOTAL: 6.340
*354 Illegal Votes Deducted from Huggins's Total
Precinct Illegal Votes Percent of Vote Rec’d Deduction
Bird Springs 2 68.4 1.368
Black Mesa 1 34.4 .344
Chilchinbeto 2 51.3 1.026
Cibecue 3 77.2 2.316
Low Mountain 1 53.0 .530
Pinon 2 38.9 .778
Whitecone 2 51.6 1.032
White River 2 63.3 1.266
8.660 VOTES DEDUCTED FROM HUGGINS’S TOTAL:

After rounding off the totals, we deduct six of the illegal votes from Patton's total and nine from Huggins’s total, leaving Patton the winner with 3,587 votes over Huggins’s 3,576.