Policy Review Banner

The Lessons of Florida 2000

Saturday, December 1, 2001

It has now been a year since the U.S. Supreme Court cut off the manual recount in the Bush-Gore election. In the heat of the battle, it seemed few could dissociate who they thought should win from how they thought the disputed issues could be resolved. Perhaps for many this is still true. But for the rest of us, the events of last year may have acquired a sufficient distance, lengthened by the intervening shock of September 11, to allow us to put down our partisan positions and ask ourselves how we would want these issues resolved in the future.

If we learned anything from this election, it was that resolving election issues in midstream, after we know which candidate will benefit from any given resolution, is a recipe for disaster. Right now, while we are still behind the veil of ignorance and do not know which candidate will benefit, we need to resolve as many open issues as we can.

Some of the issues will be familiar, since they arose in the actual dispute. But they have been greatly misunderstood, both because the rapid pace of developments did not allow for sufficient explication and because insufficient attention was paid to how the candidates would want the issue resolved if their positions were reversed in the next election. Other issues will be unfamiliar, for they are the issues that would have arisen had the Supreme Court not called a halt, and may well arise the next time around. And if anything was clear to me as I nervously anticipated those issues at the time in my role as counsel to the Florida House, it was that the issues barreling down the track would have been even more explosive and bitterly contested than the ones we all argued about a year ago.1

 

Lesson i. Better voting machines. Let’s begin with the seemingly most obvious lesson: We need better voting machines. But now the nonobvious point: The problem was not, as conventional wisdom thought at the time, with punch card technology. The exhaustive media recounts have confirmed that punch card and optical-scan ballots actually resulted in similar rates of spoilage, defined as the total of undervotes and overvotes. (An undervote is a ballot that registers no vote for a candidate, while an overvote is a ballot invalidated by votes for multiple candidates.)

How can that be? Didn’t we all at the time hear statistics that seemed to confirm the superiority of optical scanners? Well, not quite, for two reasons. First, although the focus at the time was on the undervotes that Gore and the Florida Supreme Court wanted recounted, it turns out that there were twice as many overvotes, and they were a bigger problem on the optical-scan ballots.

Second, and more important, the counties in which optical-scan ballots seemed to be delivering better results were actually only those counties that counted ballots at the precinct level. Under these systems, the ballot result is registered or rejected by the machine when the voter turns in the ballot. Such a precinct counting machine provides voters with timely feedback that allows them to correct any errors in their ballots. But when punch card ballots were also machine counted at the precinct level, they had a similarly lower rate of spoilage. When either optical-scan or punch card ballots are machine counted at a centralized county location removed from the precinct and the voter, the rate of spoilage is higher, but similar for both types of ballots.

The implication is that what really would constitute “better” voting machines are not machines better at counting, but machines that are better at correcting voter errors. This is not at all to trivialize the concern. No matter how smart we may think we are, we all push the wrong elevator button from time to time. Machines that would help us avoid these inevitable errors in voting are important and useful. But correctly understanding the issue should put to rest the misguided bugaboo that certain counting machines were disenfranchising voters. The actual problem is that locating the counting machines too far from the voters disables those machines from helping to prevent voters from disenfranchising themselves.

Should we care about this problem? Some argue that the word “spoilage” seems inapt because sometimes voters intentionally don’t vote for any candidate for a particular office because they don’t like any of them, or vote for multiple candidates because they regard the choice as too close. We thus cannot assume that under- or overvotes fail to accurately reflect the voter’s intent. Perhaps we should make “none of the above” an explicit ballot option. Still, when some voting setups regularly produce many more under- and overvotes than others, the difference is likely due to unintentional errors. So it makes sense to create machine setups to give voters the feedback that helps them minimize those errors. We should do so not only for the sake of those of us who make such errors, but because the more precise the ballot answers we can get from voters, the less likely we all will suffer through bitter controversies about how to interpret the ambiguities they leave behind.

Should we instead be worried about a different problem: the distribution of machines of supposedly poor quality? Many were concerned by the evidence in Florida that a greater proportion of minority voters had spoiled ballots, producing the suspicion that they were saddled with worse machines. But the recount undertaken by the National Opinion Research Center on behalf of a consortium of media outlets indicates the higher minority spoilage rate was true regardless of machine type, and that the distribution of each machine type was similar for black and white areas. Nor can one assume that, because most minority voters are Democrats, any disparity hurt or was targeted at Gore. A study by John Lott reports black Republican voters were an astonishing 50 to 66 times more likely than the average black to have a spoiled ballot, and that black Democratic voters actually had a slightly lower spoilage rate than the average voter. His findings also seem inconsistent with any conspiracy by white or Republican election supervisors, since the rate of spoiled ballots was 14 percent to 31 percent higher in precincts with black or Democratic election supervisors. The overall rate of spoiled ballots was also higher for white Republican voters than white Democratic voters. In any event, the lesson for the future should be to have the best machines possible in every county — those that provide immediate feedback. That decision can either be made at the state level (as Florida has now done) or by the counties, which should have ample incentive to pick the best machines for themselves since to do otherwise would merely reduce their own electoral clout.

Lesson ii. Manual recounts should be limited to cases of machine malfunction. Having deployed machines to help reduce human-created ambiguity in the act of voting, it also makes sense to deploy them to reduce such ambiguity in the counting. Manual recounts require problematic subjective interpretations of ambiguous marks left by voters, and the counters have their own political preferences that (consciously or not) are likely to bias their counting.

The nearly year-long media recount completed in November confirms the problem. This was a highly professional recount, conducted under standards applied uniformly within each county by counters who were screened for political bias. Even so, counters frequently disagreed in their ballot interpretations. Although counters agreed on 96 percent of punch card ballots, that 4 percent error rate greatly exceeded the 0.001 percent margin in the Florida presidential election. Thinking that human recounts with that sort of variation can check the accuracy of a machine count is rather like trying to recheck a machine’s measurement of electron width using the human eye and a yardstick. Moreover, this 96 percent figure is misleading because it includes agreements on ballots where there was no marking to dispute. On ballots where at least one counter saw a potential vote for Bush or Gore, the counters disagreed 34 percent of the time, 37 percent for punch card ballots.

Most worrisome, even with elaborate efforts to screen for political bias, the political affiliation of the counters affected the results. Republican counters were 4 percent more likely than Democratic counters to deny a mark was for Gore. Even more striking, Democratic counters were 25 percent more likely to deny a mark was for Bush.

None of this should be taken as criticism of the National Opinion Research Center, a highly respected nonpartisan outfit that made scrupulous efforts to check for political bias. Nor does it necessarily reflect badly on the counters, whose bias might well be unconscious. Rather, the problem is endemic to using humans to count ballots. Indeed, the lesson is quite the contrary. If this is the sort of accuracy and bias one gets from an unhurried, professional nonpartisan organization whose counters were screened for bias and bound to the same standard, imagine the sort of inaccuracy and bias that would result from a partisan set of counters, rushing to complete a recount quickly, and free to vary their standards.

But don’t machines have their own inaccuracies, failing to register votes when chads are punched out but hanging on by a cardboard thread, or when some misguided voter both fills in the box for a candidate and also casts a write-in vote for the same candidate? Sure. But in analyzing the issue we must be sensitive to relative inaccuracies. A human recount that is 4 percent inaccurate cannot improve upon a machine count that is 1 percent inaccurate. More important, even if the error rates are equal or even higher for machines, the machines are far less likely to alter the result. Even if machines failed to detect votes at an enormous rate, like 10 percent, that will rarely alter the result because machines make any errors randomly, and thus should undercount a proportional share of each candidate’s votes. A much smaller error rate that is biased toward one candidate can, on the other hand, produce a significant net change and thus alter the result. Moreover, manual recounts not only raise problems of misinterpretation and bias, but also are more likely to produce simple math errors, may alter ballots through handling, or even involve conscious mischief or fraud.

So whether my candidate is ahead or behind in the next election, I would still conclude manual recounts should be allowed only when there is some machine malfunction that gives us reason to think the machines are far less accurate than normal. And while we all don’t know whose ox will be gored, we should adopt that lesson for the future. After all, machine counting was introduced in this country not just for speed and cost, but to reduce the fraud and other human error that used to routinely attend ballot counting. Under a regime of manual counting, as Stalin said, “The people who vote decide nothing; the people who count the vote decide everything.” There is no reason to subject ourselves to such a regime when we have reasonable alternatives.

Now, permit me a detour into the partisan-tinged past that I think carries a lesson for the future. In fact, there were powerful reasons to think the Florida statute in 2000 did limit manual recounts to cases of machine malfunction. As counsel for the Florida attorney general (Gore’s state chairman) conceded, before this litigation Florida had never allowed a manual recount to be based on a claim that humans can interpret ballots better than machines. To the contrary, such claims had been repeatedly rejected, including in Broward and Palm Beach counties. Instead, the statute provided that manual recounts could be done only if the machines malfunctioned. That is why manual recounts had to be county-specific — they were supposed to rest on a defect in that county’s machines — rather than statewide as they would be if the defect were a general problem in interpreting ballots. That is also why a strict deadline was set: Since manual recounts were supposed to be mere counting exercises rather than elaborate efforts at interpretation, seven days with an ability to hire many counting teams more than sufficed. The only statutory remedy for a close election was a machine recount.

The Florida Supreme Court instead interpreted the statute to provide that manual recounts must always be done in any close election. So much attention was focused on the deadline issue that the most dubious aspect of this interpretation was missed by almost everyone: the Florida court’s conclusion that judicial contests could overturn a certified election result without showing any illegality by election officials at all. This deviated from prior caselaw requiring “substantial noncompliance” with the law, including a case specifically holding that a contest could not be used to challenge a county’s lawful decision not to conduct a manual recount to pick up partially perforated chads. The ostensible excuse was that the Florida legislature had enacted a 1999 statute on contests that changed this law. But the statute on its face required the “wrong” of rejecting a “legal vote,” not a ballot legally interpreted to cast no vote. And the legislative history stated that the statute “codifies” prior law, not that it changes the law. Indeed, in 2000, just one week prior to its conclusion that the “substantial noncompliance” test could not bar Gore’s contest, the Florida court had ruled 7-0 that another contest should be dismissed for failure to show substantial noncompliance with law.

The lesson for the future from this retrospective detour lies in the question: What drove the Florida court to this contentious interpretation? Republicans naturally suspect partisanship. After all, the Florida court was dominated by activist liberal Democrats, most of whom had been chosen for appointment by Gore’s counsel. But the deeper reason was probably the court’s deep-seated empirical premise — never critically examined — that manual recounts are more accurate than machines at interpreting the intent of the voter. Future courts are likely to share this premise regardless of partisan considerations, because courts as a class tend to have an exaggerated view of the accuracy of their own processes and thus have a hard time saying there is any wrong (such as some inaccurately counted ballots) without a remedy. Herein lies the lesson: When legislatures restrict manual recounts, they must be extraordinarily clear. Any such statute needs to state explicitly why and when it regards such a judicial cure as worse than the supposed machine-caused disease.

 

Lesson iii. Adopt objective standards in advance to constrain human decision-making. A third lesson flows obviously from the prior analysis: To the extent humans make post-election decisions that affect who will win an election, they must when possible be governed by clear objective standards established in advance of the dispute. We cannot allow humans, who inevitably have their own partisan views, to exercise standardless discretion in the midst of an election crisis when they and everyone else know which candidate will benefit. The risk is simply too great that any decision they make will be biased (consciously or unconsciously) in favor of their preferred candidate, and even if uninfected by actual bias will be perceived to be biased in a way that taints the legitimacy of the result. In the 2000 election, the big issues were about whether, where, how, and which ballots to recount manually, but the basic proposition seems generalizable to any issues left to midstream resolution by humans.

I say this lesson is obvious, and yet the United States Supreme Court has gotten a lot of undeserved grief for effectively imposing this norm under the equal protection clause. Critics argue that Florida’s manual recount process — while inaccurate, arbitrary, and haphazard — was not unconstitutional. Ronald Dworkin, for example, argues that the equal protection clause is only violated when state law creates “distinctions that put some citizens, in advance, at a disadvantage against others.” But what made this process alarming was precisely that it did not set forth any objective standards “in advance.” Such standardless discretion in the hands of partisan county officials is worrisome because it allows them to engage in sub rosa discrimination against the opposing party about how (and indeed whether) to conduct manual recounts. Since without standards such discrimination is hard to prove, the best way to vindicate the constitutional right of equal treatment is to prevent the partisan officials from exercising such standardless discretion at all. For precisely this reason, well-established Supreme Court precedent makes standardless discretion unlawful if used to hand out parade permits or locate newspaper boxes. Why should the protection be any less when discretion is being exercised over the far more fundamental question of which votes to count?

This understanding of the Supreme Court decision also explodes another critique: that it is internally incoherent because if one really accepted the court’s logic, any election in which some counties use better voting machinery than others would also violate the equal protection clause. But this misses the point that the concern was not with county differences, but with standardless discretion that would allow discrimination against candidates. Just as no constitutional difficulty is raised when different counties in advance set forth different hours for parade permits, so too no worry about sub rosa discrimination is raised when counties have different voting systems as long as they are adopted in advance. No county has incentives to reduce its own clout, so any decision it makes is essentially a judgment about which machines count best or a tradeoff between that goal and the costs of new machines. Different counties may make different judgments or tradeoffs, but as long as they make their judgment in advance, it does not reflect one party trying to manipulate the electoral rules to discriminate against the other party. What is worrisome is when partisan election officials get to decide whether and how to do the recount after they find out which party is a little behind.

The same goes for setting standards for interpreting ballots. If, in advance of an election, counties adopt differing standards for interpreting punch card ballots — one a “two corners” standard, and another a “four corners” — that should raise no constitutional difficulty because no county has incentives to adopt an interpretive system that undercounts its citizenry. It is when county officials make such decisions knowing which candidate will benefit that the problem is raised. As the media recounts emphasize, one gets a different vote count for every standard one chooses. But as long as the counties choose their objective standard before they know who will benefit, they should be free to adopt differing standards. A dimple standard should probably be rejected given the evidence from the media recounts about how particularly variable (and subject to bias) human judgments about dimples are. But any objective standard that reasonably constrains human discretion should be permissible.

A critique that is more on point argues it cannot be unconstitutional to apply an “intent of the voter” standard because similarly vague intent standards are applied all the time by juries. But, as the U.S. Supreme Court explained, such vagueness is allowed only for issues that are not susceptible to more objective standards. Rules for counting ballots can easily be more objective — one need only pick an objective standard before the election and stick to it. The Supreme Court is not asking the impossible, but when an objective standard is possible, state officials cannot instead retain a discretion that maximizes their influence over results.

A more legalistic objection, offered by my colleague Alan Dershowitz, is that past equal protection caselaw has required proof of facial or intentional discrimination. But this objection does not really come to grips with the caselaw on standardless discretion, where the very objection is that the lack of any standards against which to test the results makes it impossible to prove intentional discrimination, and thus requires the imposition of standards to root out sub rosa discrimination. Nor does the objection accurately describe modern election law since the leading case of Anderson. That case was brought by presidential candidate John Anderson in the 1980 election, challenging the application of a state’s filing deadline to his decision to run as an independent after he lost in the Republican primaries. The filing deadline was facially neutral and no showing was made that it was intentionally discriminatory. Nonetheless, the Supreme Court struck it down on the grounds that it was set too early for independent candidates. Instead of requiring facial or intentional discrimination, the court articulated the now-prevailing test that state election law must in its effects be both nondiscriminatory and reasonably related to important state interests. It is no great leap to say the imposition of a system of standardless discretion for interpreting ballots does not meet those requirements.

A related objection is more visceral: Manually counting ballots is too traditional to be unconstitutional. After all, for over a century that was the only way of counting ballots, and for the remainder of our history manual recounts have been accepted. But traditional practices often become unconstitutional: Indeed, the recognition of a constitutional right to an equal vote itself overturned nearly two centuries of contrary practice. The invalidation of traditional practices is particularly appropriate when changed circumstances alter their implications. The sort of filing deadline at issue in Anderson, for example, also had a long and accepted history. But it was invalidated nonetheless, with the court reasoning that the “passage of time since the Constitutional Convention in 1787” had changed the deadline’s reasonableness because changed technology and expanded literacy made spreading information about late-developing independent candidates more feasible. Here, too, changed technology is relevant, making it both more necessary and possible to limit human discretion in ballot interpretation — more necessary because machine ballots leave things like partially perforated or dimpled chads whose significance is much harder to interpret than a mark next to a candidate’s name on an old-fashioned paper ballot; more possible because rules more precise than voter intent can be developed for determining how many chad corners must be perforated, and because the machines themselves are always available to count the ballots free of concern about human bias and subjectivity. In any event, the Supreme Court never outlawed all manual recounts, just those conducted without any consistent standards.

The above objections also suffer from being fixated on what rule governed in the past. The important question now is what rule should govern the future, and it is hard to see a convincing reason why we wouldn’t want a rule that limits standardless discretion by election officials in future elections. True, Alan Dershowitz argues the real problem is partisan decision-making by the U.S. Supreme Court. He claims there is no possibility the court majority would have reached the same conclusion had the position of the candidates been reversed. But seven of the nine justices found an equal protection violation, including Democratic Clinton appointee Stephen Breyer. And it does not seem at all implausible that they would have reached the same conclusion if the party affiliations of the candidates were reversed, and that they would and should do so in the future.

If you are a Gore supporter, consider the following thought experiment. Suppose in 2004 Al Gore runs against Bush again and the election again hinges on a close result in a single state, only this time Bush is slightly behind after the machines count the votes. Bush seeks a manual recount only in a selection of highly Republican counties. State law provides for machine counts and recounts, and allows manual recounts only for county-specific errors in machine tabulation, not based on a claim that humans can interpret ballots more accurately than machines. Bush’s counsel concedes manual recounts have never been used for this purpose under the state statute, and the Democratic secretary of state vested with authority to interpret the statute interprets it not to permit manual recounts for those purposes. Consistent with this, the Republican county boards initially deny manual recounts, but later reverse this decision under pressure from national Republican leaders working for Bush. The Republican county boards then assert that, even without any machine defect, they have unfettered discretion to decide whether to do manual recounts when election results are close. They are upheld by a state supreme court dominated by highly activist conservative Republicans, most of whom were appointed by Bush’s counsel. The Republican county boards begin the recount under the pre-existing perforation standard, but switch standards when the first one produces little net gain for Bush. When this still does not create much of a net gain for Bush, they switch to a policy of exercising discretion over which standard to apply. When the counties fail to finish the manual recount by the statutory deadline, the Republican state supreme court extends it. When the counties also fail to meet the new judicial deadline, the Republican state supreme court holds that judicial contests can overturn election results without any legal violation despite longstanding precedent to the contrary, including a case the same court had decided a mere one week earlier. Although the Republican court had earlier been willing to allow Bush to pursue selective manual recounts in heavily Republican counties, now that it has become clear such a selective recount cannot produce enough Bush votes, the court orders a statewide recount that Bush never requested. Partisan county election officials are left with complete discretion over what standards to use, and Bush’s counsel concedes the standards being applied vary not just from county to county but from table to table.

Would not Gore supporters in such a case feel justifiably aggrieved? And do we really have any doubt that in such a case the U.S. Supreme Court would — and should — reach the same conclusion it reached in the actual case? Contrary to Dershowitz’s claim, I think the vote would have been — and would and should in the future be — unanimously for Gore in such a case.

Still, critics complain, the proper remedy should have been to remand for a recount done under uniform standards that constrain discretion, not to stop the recount altogether. Now I must admit that at the time this critique seemed forceful: Equal protection claims naturally suggest the remedy of redoing things equally. But on reflection, the Supreme Court did the right thing, for reasons that are likely to arise in the future and should produce the same result no matter which candidate benefits.

First, the Florida Supreme Court had already held — as any state court would likely hold in a future case — that under state law the time for finishing presidential contests must be circumscribed by the deadline federal law sets for making the resolution of judicial contests binding on Congress when it counts electoral votes. True, in a normal case the U.S. Supreme Court would simply remand and leave such a state law issue to the state supreme court, in part because the state court might change its view of state law.2 Having changed the state deadline on manual recounts twice before, it would hardly have been surprising if the Florida Supreme Court had tried to do so a third time. But the situation with presidential contests presents a special case because (unlike in other areas) federal law constrains the power of a state court to change its understanding of state law under Article II, which imposes a constitutional requirement to follow the state legislature’s directions. Accordingly, a state court that says one week the state legislature’s directions imply one deadline for presidential contests cannot the next week change that deadline without violating the federal Constitution. Thus, this is one case where it makes sense for the U.S. Supreme Court to express its view about the state law baseline. Had the luxury of time been available, the U.S. Supreme Court could have first remanded without expressing its view and stood ready to reverse under Article II if the state court tried to change state law. But since its opinion came out on the eve of the federal deadline, the U.S. Supreme Court had a compelling reason to express on that date its understanding of what the Florida Supreme Court had said about the state law deadline — to create a judicial resolution in time to assure Florida’s electoral votes would be counted in Congress, as the Florida Supreme Court had said it wanted and state law required. (It also staved off a state legislative appointment of electors that would have occurred the next day had the federal deadline passed.) In effect, the U.S. Supreme Court’s statement adopting the Florida court’s prior statements about the Florida deadline for presidential contests telescoped into the present any future Article II decision that would have been necessary had the Florida court tried to change that deadline.

The second reason is more directly connected to the underlying equal protection theory. To really satisfy the objections to using standardless discretion to resolve election issues, rules that constrain that discretion must be adopted in advance. Adopting objective standards in midstream does not help if by then the decision-maker knows who will benefit. Had the U.S. Supreme Court remanded for the Florida Supreme Court to pick a standard, that court would have had to exercise standardless discretion about which standard to use. Perhaps early in the process, the Florida court could have picked the perforation standard that at least was the pre-existing test in Palm Beach. But by December 12, both high courts knew too much. The partial recounts and revealed preferences of Bush and Gore indicated that a perforation standard favored Bush and a loose dimple standard favored Gore. Choosing the standard was thus tantamount to choosing the president. Nor was there any objective test for making a choice among the standards for interpreting ballots. Either would be constitutional if chosen in advance. But to choose among them in midstream, knowing which candidate would benefit, would be to exercise the very sort of standardless discretion that made the manual recount unconstitutional. Remanding thus could not cure the constitutional problem; it would simply have made the Florida justices, rather than county commissioners, the state actors exercising standardless discretion. Given this dilemma, the best solution is to stick to the only objective standard that had been picked in advance — the interpretation provided by the machine count — which is effectively what the U.S. Supreme Court did by stopping further manual recounts, and what it should do again in a similar future case no matter what the identity of the candidates.

Despite what Gore repeatedly said in the last election, the dispute was never about whether to “count every vote.” The issue was — and will be in the next dispute — about how to count the ballots. The U.S. Supreme Court correctly established that counting methods cannot involve midstream exercises of standardless discretion. They must instead conform to objective methods established beforehand by lawmakers who are behind a veil of ignorance about the question of who benefits. It is when we are behind that veil for the next election, and don’t know which candidate will benefit from our rules, that standards should be set.

This is hardly to say the problems were all on one side. Well before the election, the Florida legislature had directed the Florida secretary of state to “adopt rules prescribing standards” that “shall ensure that ballots are counted in a uniform and consistent manner.” Had Katherine Harris obeyed this statutory mandate, the entire mess would likely have been avoided. George Bush had signed a bill in Texas allowing for the counting of dimpled ballots in a way that also probably fails to provide sufficient protection against standardless discretion. But the point is not to assign blame for the past, it is to preclude similar problems in the future. Now that the U.S. Supreme Court decision in Bush v. Gore is the law of the land, it is incumbent upon every state to bring its election law into compliance by minimizing whenever possible any role left to standardless discretion exercised by humans.

 

Lesson iv. Limit partisan involvement in election decisions. No matter what scope is given to machines, and how well-defined our advance standards, some irreducible role will be left to human decision-making in resolving election disputes. It will thus be important to heed the fourth lesson from the 2000 election debacle: Those who make election decisions should be rendered as non-partisan as possible.

At a minimum, election officials should not serve as campaign chairs for any candidate. It is tempting to do so because elections are not normally close. But being an election official is a job that only really matters in close elections. Here, both sides were at fault in Florida. Both Secretary of State Harris and Attorney General Bob Butterworth were the respective state campaign chairs for Bush and Gore. But Harris’s conflict of interest proved the more damaging. Her initial interpretation that manual recounts were not authorized absent machine malfunctions was actually completely right on the law, and if respected would have nipped the dispute in the bud. But her clear conflict of interest made her all too easy to ignore. No election official should be in a position of being disabled from giving authoritative neutral advice when it matters most. This was the norm in most states already, but it should be codified and extended to all states.

I would go even further. In the Florida election, it seemed plain that election officials were influenced — or at least could justifiably be perceived to be influenced — by the political threat posed by the need to get re-elected, often by narrow county constituencies. It also seemed clear both election officials and judges suffered in their legitimacy because they were constantly described by their political affiliation, as in the “Democratic Florida Supreme Court” or the “Republican U.S. Supreme Court.” It is difficult to have confidence in election resolutions under such a system.

To address these problems, election officials should be appointed to insulate them from political influence during elections. Further, both they and judges, as a condition of taking office, should give up their party affiliation and right to vote in future elections. There is some precedent for this. For just these reasons, Justice John Marshall Harlan refused to vote in presidential elections, and Canadian law prohibits judges from voting. To be sure, this is no panacea. Any election officials or judges will inevitably have been appointed through some political means and be identifiable by the partisan affiliations of the officials who nominated them. But it should be understood that just as a monk gives up earthly possessions when entering the monastery, so too election officials and judges give up their partisan affiliations when they take on their roles. They should not have a rooting interest in elections. We should not have Democratic or Republican election officials or judges; we should have election officials or judges for all of us. Giving up their political affiliation and participation would be one important way to signal that commitment.

 

Lesson v. Define the rules for state legislative involvement. Now we come to the even more controversial set of issues that would have arisen had the U.S. Supreme Court not called a halt to the manual recounts. Of these, perhaps none is more explosive than defining the role of state legislatures in directly appointing presidential electors when elections go bad. The Florida House had already voted to appoint electors before the U.S. Supreme Court ruled on December 12, and the Florida Senate was scheduled to complete the appointment on December 13 if the court failed to resolve matters. Then all hell would have broken loose, because Gore supporters vociferously disputed the power of the state legislature to make such an appointment. And there are some reasonable grounds for disagreement that we should, again, resolve now — rather than wait until we are in the midst of another bitter election dispute and know who might benefit.

A bit of background is necessary. The Constitution requires that presidential electors shall be chosen in whatever manner state legislatures direct. Early in our history this was often done by direct legislative appointment, but since then legislatures have fallen into the salutary democratic habit of directing that electors be chosen through state elections conducted according to legislatively prescribed rules. Congress has two relevant constitutional powers: the power to set the time for the choosing of electors, and the power to count the electoral votes. Pursuant to the first power, it has promulgated a statute providing that if a state’s election “has failed to make a choice” on the election date prescribed by Congress, then the state legislature can after election day appoint electors in any manner it deems fit. Pursuant to the second power, it promulgated a statute providing that, in deciding whether to count electoral votes, Congress will regard the results of a state judicial contest as binding if it is completed six days before the Electoral College meets (which in 2000 meant December 12) and is resolved in accord with pre-existing law.

Unfortunately, neither the statute nor any caselaw provides any criteria for deciding when an election “fails to make a choice” or who gets to decide when no choice was made. Nor had any case addressed who has the power to decide when a state court has failed to comply with pre-existing law and legislative directions. As counsel to the Florida House of Representatives, I took the view that if the election failed to make a choice conclusive on Congress, it should also be deemed to have failed to make a choice conclusive on the Florida legislature, which should thus be prepared to directly appoint the electors. If the election contests had not been resolved by December 12, they clearly would have failed to make a choice conclusive on Congress under the federal statute, and the legislature would have been justified in directly appointing electors to assure Florida was represented in the Electoral College. I also thought that, since the state legislature was the one responsible for directing how electors were chosen, it had the responsibility for determining whether its election had failed to comply with pre-existing law or to follow legislative directions in a way that satisfied constitutional requirements.

These views were controversial. Others, most notably Bruce Ackerman, took the view that an election did not fail to make a choice unless no choice was certified by the state executive or state supreme court. The fact that the election choice might not be conclusive on Congress was, to them, irrelevant. They further thought that the question of whether election contests complied with pre-existing law was a matter of state law on which the final authority was the state supreme court. Based on such arguments, the Gore forces were preparing lawsuits seeking a court order blocking any appointment of electors by the Florida legislature. There was talk of a court order prohibiting the Florida legislators from meeting to make the appointments, or barring any legislatively appointed electors from voting at the Electoral College. Given that the Electoral College had to meet on December 18, this would have left but a few short days to litigate and resolve all appeals concerning the complex and never-before-adjudicated issue of whether state legislative appointment was appropriate when election contests failed to follow state legislative directions or make a timely choice conclusive on Congress.

Now, when matters are less frenzied, would be a better time to take a more considered look at the issue. Since neither the statute nor caselaw provides any direct answers, we must thus turn to legislative history and structure. Those who argue that the “failed to make a choice” language only applies when the election literally produces no result point to legislative history indicating Congress was concerned about circumstances in which state election law requires a majority and no candidate has received one. But there was other legislative history indicating that Congress was also concerned about instances in which floods or inclement weather prevented “any considerable number” of voters from reaching the polls, and that Congress wanted to confirm the power of the state’s “legislature to authorize the continuance of the elections” past the congressionally prescribed election day. The latter legislative history makes clear that an election might “fail to make a choice” even though there had been an election with a result, at least when that result was distorted by flooding or bad weather. It also makes clear that, at least in that circumstance, Congress contemplated that the state legislature was the entity that would decide whether the election had failed to make a choice. Unfortunately, the legislative history does not indicate what else Congress thought might make an election fail to make a choice. But at a minimum, this legislative history seems to rebut the “no certifiable result” limit suggested by Ackerman and others. Further, one might reasonably conclude by analogy that the state legislature should have the power to decide when in its judgment other problems created distortions in the election result.

Statutory and constitutional structure suggest the same result. A standard canon of statutory construction requires reading different statutory provisions together to make a coherent whole. While the limits on when a state legislature can say that an election failed to make a choice are unclear, it makes little sense to have one provision say that certain election results fail to make a choice conclusive on Congress, and then interpret another provision to have those same election results make a choice conclusive on the state legislature. This does not seem a plausible reading of a federal statute that was, after all, merely intended to regulate the timing of elections. Nor does certification by some state official eliminate concerns about whether a state’s electoral votes will make a choice that will be counted. Rather, the federal statute provides that election results are not binding on Congress when it counts electoral votes unless any “controversy or contest concerning the appointment of all or any of the electors” has been finally determined in conformance with pre-existing law before the statutory deadline. A certification of appointment thus does not suffice to make a binding choice if it is challenged by a pending “contest.” Moreover, even without this statute, a certified choice could well be invalidated as unconstitutional if it violated equal protection or the constitutional requirement that elections be resolved according to the state legislature’s directions.

Further, another canon provides that statutes should be read to avoid constitutional doubts, and even without that canon it makes sense to read statutes consistent with the constitutional structure. As courts have repeatedly confirmed, the Constitution gives state legislatures plenary power to appoint electors. It would be in considerable tension with this power to interpret the statute to mean that Congress has said to the states, in effect, “We may not count your electoral votes, but there is absolutely nothing you can do about it.” I doubt such a statute would be constitutionally valid at all, but there is certainly no reason to think Congress intended such a meaning. Moreover, the whole reason to insist that pre-existing law be followed is to comply with the constitutional requirement that electors be chosen in compliance with the state legislature’s directions. Who better to decide whether its directions have been followed than the direction-giver itself?

In truth, the objection to state legislative appointment rests less on legal analysis than a policy objection that it would invite frequent and destabilizing legislative intervention that is undemocratic and illegitimate, turning us (as Ackerman charged) into a “Banana Republic.” But neither the empirical claim about frequency nor the normative claim of illegitimacy seems well-founded.

In the 113 years since the Electoral Count Act was enacted, Gore was the first losing presidential candidate to contest an election at all. In 113 years, we had not had contests that lasted beyond the statutory deadline set by Congress for making the choice produced by those contests binding. In 113 years we had not had a state supreme court render an opinion that a unanimous U.S. Supreme Court thought (recall its first opinion) could reasonably be read to both circumscribe the state legislative power to direct the manner of appointments and raise the risk that Congress might deem the state supreme court to have effected a change in law that would deprive the election results of their conclusive binding effect. And in 113 years we had not had that same state supreme court render a second opinion that its own chief justice thought had “no foundation” in pre-existing state law, and that at least three (and probably five, judging from oral argument) justices of the U.S. Supreme Court thought failed to obey the constitutional requirement of following legislative directions. There seems little reason to think such an occasion will arise repeatedly. Indeed, it did not ultimately result in 2000 because the U.S. Supreme Court ended the contests in time and brought the results into conformance with pre-existing law.

The argument to the contrary seems based less on empirical information than on an assumption state legislatures will find it irresistible to try to overturn election results even when no legitimate controversy about that result exists. But legislatures will not want to overreach for the same reason that legislatures have not taken back the power to appoint generally: Legislatures are reviewable by the electorate. State legislators know that if they try to overturn the defined choice of the state’s voters, they will quickly find themselves out of office. That is why the Florida legislature held back on making an appointment based only on the disputed claim that the Florida Supreme Court had deviated from pre-existing law, waiting until a time when the contests would have indisputably exceeded the statutory deadline.

On those rare occasions when the election process has failed to provide a timely or convincing result, it is perfectly legitimate for the problem to be resolved by the entity that both (a) was given clear constitutional authority over the matter; and (b) is most responsive to the will of the Florida electorate. This might seem inconsistent with the prior lesson that election officials and judges should be made nonpartisan, but it is not. Those officials and judges have the job of resolving election disputes in a timely fashion and in accord with pre-existing rules. For that task, one needs a neutral umpire. But when they fail in that task, or where there are no constitutionally valid pre-existing rules, then the democratic choice of the electorate will be unclear. At that point, we should want the best proxy for the electorate’s choice, and that proxy is the institution most responsive to democratic will in any state, its legislature.

Note also that a state legislature need not proceed by direct appointment. It can cure election unclarity by prescribing any method of selecting electors it wants. In Florida, for example, the state legislature might have filled in the vacuum by ordering another election or directing that manual recounts be conducted according to a uniform standard, most likely the perforation standard that was the only pre-existing written standard.

Finally, we must realize that eliminating the state legislative role would not eliminate the concern about overreaching by government officials tempted to try to overturn the result in any close election. It would instead leave us with the concern that state judges might overreach in precisely the same way. The latter seems, if anything, more problematic because courts are not reviewable by the electorate. Thus, unlike legislators, if judges try to deny the electorate their democratic choice, there is no way for the electorate to punish them.

If the state legislature can decide when state judicial contests have left it with an election that failed to make a choice, there is the separate question of whether the U.S. Supreme Court or Congress can also make such a determination. At the time, I took the position that the matter should not be deemed justiciable. But I must admit I am no longer so sure. The experience made it clear that many legislators (both in Florida and Congress) were extremely reluctant to get involved in such a controversial dispute. Most did not want to touch it with a 10-foot pole, because they realized that, no matter what they did, they would alienate half their electorate. The bitter denunciations of the U.S. Supreme Court suggest they were right. There is thus sound reason to have some national entity whose job it is to resolve disputes others want to duck, and in our nation that entity is the U.S. Supreme Court. The prospect of review by that court, as well as by Congress when it counts the electoral votes, should also deter and police overreaching by state legislatures, thus ameliorating the concerns noted above.

I am thus inclined to the view that either the state legislature or the U.S. Supreme Court ought to be able to decide when an election has failed to make a choice. But even that leaves many questions open. Can the high court or Congress also review a state legislature decision on whether an election failed to make a choice, and if so, with what standard of review? The main lesson is that these issues should all be resolved now, when we are not driven by partisan concerns. If we as a nation decide state legislatures should have no role after election day, so be it — but we need to adopt clear constitutional or statutory rules in advance. Likewise, if we think state legislatures should have a role, the bounds and reviewability of that role need to be defined carefully beforehand.

 

Lesson vi. Clarify certification authority. Whether or not state legislative involvement is permitted, there is another issue that would have arisen had the 2000 election dispute continued, one also likely to arise someday in the future: What do we do when state officials render conflicting certifications?

The Electoral Count Act vests each state governor with the authority to certify the winner of the state’s electoral votes, a certification that can break deadlocks when the two houses of Congress reach different conclusions about which of two electoral slates to count. Pursuant to that authority, Governor Jeb Bush had already certified electors for his brother George based on the election results before the judicial contest started. But if the Florida court had pushed through a recount that made Al Gore the winner, could it then order Governor Bush to certify the court’s slate and decertify his original slate? It is not at all clear whether the state court has any authority to interfere with the certification authority that the federal statute vests in state governors. Nor was there any precedent on whether it is a proper exercise of such certification authority for the state governor to certify only election results provided by pre-existing election law and thus refuse to certify the result of a judicial contest that, in his view, was produced by changing the rules in midstream.

Further, if the state court entered such an order, what would happen if the governor refused to follow it? One wonders whether court marshals would be dispatched to fight their way through state troopers and seize the governor in order to jail him until he signed a document decertifying the Bush electors and certifying the Gore electors. Perhaps the governor would avoid seizure by taking a trip out of state and beyond the jurisdiction of the state court. Or perhaps the governor would just sit in jail and refuse to sign any change of certification until after the electoral votes were counted.

It seems best to have clearer rules set in advance to assure the next election does not turn on such drama. The Electoral Count Act, by giving effect to both gubernatorial certifications and judicial contests, exacerbates the problem. It seems fairly easy to instead structure power in a way less likely to provoke such controversies. If governors are the final arbiters, then their decisions should not be reviewable by courts. But if the courts are to be the final arbiters, then they should be given the power to make the certifications so that they need not act through governors. Given that the courts must already play a role in conducting contests, I am inclined to the latter view, and would thus give the state supreme court the power of certification. If the positions above were adopted, that certification would remain reviewable by the state legislature and the U.S. Supreme Court, but if they were rejected I would just leave the state court as both the final arbiter and certifier.

 

Lesson vii. Clarify congressional counting rules. Although the Electoral Count Act provides elaborate counting rules, they are unclear in many respects. Under those rules, a lot turns on whether Congress has one or two election returns from a state before it. If a state only has one return, then it must be counted unless both the House and Senate agree to sustain an objection to counting it. Thus, if they disagree, a single return must be counted. If Congress has two returns before it, the two houses acting in concert can choose which one to count, but if they disagree the gubernatorial certification decides which slate counts. However, the act nowhere specifies who can submit a valid election return. Presumably this is not an option open to any citizen with pen and paper. Can a state court submit something that counts as a return? Could a state secretary of state submit a return that conflicts with the governor’s? The state legislature? The candidates themselves? Members of Congress? Rather than allow for multiple returns, I would as discussed above identify one entity with the power to render a certified return with any effect in Congress. But if multiple returns are permitted, some limits must be defined on who can render them.

Another unresolved issue that loomed large in the 2000 election was what to do about conflicts of interest. Assuming a party-line vote, the only way the two houses of Congress could have disagreed would have been if Gore and Lieberman had cast the deciding votes for themselves. Having the election turn on a decision involving such a blatant conflict of interest would have been disastrous. But it has constitutional precedent. Back when the Constitution authorized the sitting vice president to personally decide all disputes about whether to count electoral votes, John Adams had no problem resolving every single dispute in his own favor in order to ensure his election as president. But times have changed, and today such a spectacle would surely undermine the legitimacy of election results. Moreover, if there were two slates before Congress, the spectacle of these candidates casting the deciding votes to adjudicate the dispute in their favor would have been trumped by the spectacle of the other candidate’s brother casting the deciding certification. Now, when we don’t know who will be in what position, is the time to adopt conflict-of-interest rules to preclude such unseemly results.

Finally, we need to rediscover an obscure distinction about whether a president needs a majority of the “counted” or “appointed” electors. Suppose a single Florida slate had reached Congress and got thrown out because the two houses of Congress sustained an objection to its validity, or (perhaps more likely) the slate was invalidated in court? This might have happened had the only slate of electors before Congress ended up being one chosen pursuant to the Florida court’s manual recount, and had a federal court later (perhaps after December 18) invalidated that slate on grounds that the recount violated the equal protection clause. That would have left Congress with no valid slate of Florida electors to count. What would have happened then?

This worried me a lot at the time because I wanted to assure Florida was represented in the Electoral College, and nothing struck me as more perverse than the prospect that Florida might endure this whole fiasco and not get counted at all. It did not, however, seem to worry Democrats much. They figured that if, for any reason, Florida electors were not counted, Gore would win because he would have a majority of the rest of the Electoral College. The former solicitor general, Walter Dellinger, was advising Gore that he would win because the Constitution says you need a majority of the “appointed” electors. This was supported by precedent in cases where one state did not appoint electors and the president was selected based on having a majority of those who were appointed.

But here comes the obscure distinction. For while the exclusion of Florida electors would have given Gore a majority of the “counted” votes, he would not have had a majority of the “appointed” electors, and the constitutional text requires the latter. Further, deep within the recesses of the Congressional Record there was precedent indicating that in past cases where a state did appoint electors, and Congress sustained objections to the appointments and declined to count those electors, the president was required to have a majority of all those appointed, not just of those electors Congress counted. So while there would have been an argument, the exclusion of Florida electors most likely would have resulted in a declaration that no one presidential candidate had a majority of the Electoral College. In such a case, the Constitution provides that the House of Representatives must select the new president — which meant Bush would have been elected anyway. More important, in the future we should understand that any president needs a majority not just of the counted electors but of all the appointed electors, which helpfully reduces the incentives to try to disqualify a state’s only electoral slate.

 

The above are just some of issues that arose or would have arisen had the U.S. Supreme Court not stepped in when it did. What still leaves me worried is just how many of these issues remain unresolved and likely to arise again in any serious close case. Only some of them surfaced during the 2000 election dispute. The rest remain hidden in the deep sea of legal obscurity, ready to bob up again in the next close election.

The time has come, I think, to focus less on recriminations about the past and more on making provisions for the future. Behind the veil of ignorance, we can resolve the open issues free of the partisan jockeying that results when we know who will benefit from any given resolution. I have suggested particular resolutions to these open issues, and you may disagree with many. But I hope we may all agree that on each issue we should choose some clear rule in advance and stick to it later no matter whom it turns out to favor. Better to pick the wrong rule but have some rule we all agree to follow, than to be making up the rules in the next bitter election dispute as we go along.


Notes

1 In developing these points, I am grateful for comments on an earlier draft of this article from Randy Barnett, Richard Epstein, Richard Fallon, Barry Friedman, Mike Klarman, Nelson Lund, Dan Meltzer, Frank Michelman, Rick Pildes, and Eric Rasmussen.

2 Actually, one can question the assumption that the laws governing presidential elections involve state rather than federal law. Although adopted by state legislatures, such laws are not enacted under authority vested by the state constitution but under authority vested by the federal Constitution directly on state legislatures. This is why such presidential election laws need not be presented to the governor nor comply with other state constitutional laws. Further, the laws deemed to be federal laws under the supremacy clause are not defined as laws enacted by Congress but laws enacted “in pursuance of” the federal Constitution, which would seem to include any law enacted pursuant to a constitutional power. Likewise, the federal Constitution defines the “federal judicial power as extending to all questions arising under this Constitution, the Laws of the United States . . . or which shall be made, under their Authority.” This seems to include in the definition of a federal question any law made under the authority of the federal Constitution. Accordingly, one might conclude that when adopting a presidential election code any state legislature is exercising federal powers and adopting a federal law whose ultimate interpretation properly rests with the U.S. Supreme Court, not state courts. But that line of analysis was never pressed and played no role in the U.S. Supreme Court’s actual decision. Perhaps, though, it would help ameliorate the problem in the future to, under this theory, make the U.S. Supreme Court available to directly review wayward lower court interpretations of presidential election codes.