The Only Fair Way: What the rule of law should look like.

As the torrid electoral dispute between Bush and Gore works its way to its conclusion, anxious citizens have been forced to reflect on the proper balance between a just substantive law and the workable procedures needed to implement it. The refrain of the Gore forces employs the rhetoric of aspirations: Let justice be done; let every vote count; let the true intention of each voter be determined. The Bush camp offers the opposite approach: Let consistent rules govern election outcomes; let simple tests determine what counts as a vote; and let the principle of finality govern. The time is therefore propitious to step back from the present political brawl to ask how a well-run legal system routinely reconciles these twin demands. Aspirations would always prevail if the operation of a legal system were costless and error-free; but it obviously is not. How, therefore, must we temper these aspirations-and minimize the administrative imperfections at the same time? The oft-neglected virtues of process and formalism must come into their own.

One bedrock legal principle condemns retroactive changes in the rules. Even if the older and established rules could easily be inferior to some proposed alternative, the parties to an ongoing dispute still have the right to rely on them. If the law, for example, says that a will shall be valid once it is signed, then it should not be invalidated because it was not notarized; if the law requires the will to be both signed and notarized, then it should not be counted valid if signed but not notarized. The enforcement of existing law must be kept separate from law reform; legislative changes should be applied only to future transactions. The common saw "Don't change the rules in the middle of the game" represents a powerful bulwark against bias, opportunism, and favoritism.

The law's concern with formal virtues is also illustrated by the law of contracts. It is generally accepted that private contracts should be performed in accordance with the mutual intentions of the parties. But how are those intentions to be determined once the parties start to squabble and each side has an incentive to shade the truth or even lie? A judicial search for unexpressed intentions would quickly degenerate into a hopeless swearing contest. So the law typically allows only outward expressions of intention into evidence, even if philosophers rightly praise the subjective nature of intention. A buyer cannot ordinarily back out of a deal by saying that he meant to buy the seller's horse, and not his cow, for $1,000. Each party is entitled to rely on what the other says, so long as the stated intentions look reasonable from the context. It is too dangerous to let the seller back out of the deal if the value of the horse drops to $500.

In the same vein, the parol evidence rule normally excludes oral evidence offered to vary or contradict the terms of a written agreement. Written agreements are a good way of avoiding "he said, she said" disputes, and courts wisely refuse to open the floodgates by letting in oral evidence-which is, in general, less reliable than the written evidence it is said to supplement or contradict. Indeed, sophisticated parties often stipulate that their written agreement counts as their entire contract, no matter what oral communications have passed between the parties. Similarly, the fraud statutes in every state provide that certain contracts, such as those for the sale of real estate, are enforceable only if done in a writing signed by the party to be charged. This rule helps avoid endless wrangles over when negotiations did-or did not-ripen into a binding agreement.

None of these formal requirements is meant to restrict the freedom of contract; their purpose is to force contracting parties to create a careful record at the outset, to reduce the stresses of litigation later. No one knows how to dispense with these forms in cases where the true intentions of the parties diverged from what they said or wrote. To allow systematic breaches of these formal rules is to undermine the security of exchange indispensable to a market economy. But the conscious price we pay for our devotion to formalism is an unjust outcome in the occasional case: A rogue may escape from honoring an oral promise to sell his horse. But in the long run the formal rules work-by giving all people the incentive to say (and write) what they mean, and to mean what they say (and write). Sensible formal precautions thus reduce the frequency of misunderstanding.

Paradoxically, then, formalism increases the overall reliability of the system, by eschewing futile efforts at administrative perfection. Any inquiry into subjective intention will make more mistakes, at greater cost, than the formal rules that it displaces. Yielding to the temptation to measure subjective intention will increase in practice both the rate of error and the cost of adjudication.

The law that prizes formality also prizes finality, through the vital procedural rules of res judicata: Disputes are litigated on their merits once, not twice. Of course, the first round of litigation on the merits could have resulted in error. But that possibility hardly justifies the delay, cost, and uncertainty of a second trial, whose outcome is likely to be no more reliable than the first-and perhaps less so, if the passage of time has dulled memories and degraded evidence. The exceptions to the rule are in general very narrow; and only rarely does newly discovered evidence come to light after the first trial is completed. As with rules of form, res judicata induces people to make their best shot first, and then requires them to live with the result. If a litigant misses his most promising line of attack or defense the first time through, too bad. Over the long run, we have fewer trials-and they are better handled-under res judicata than we would in a system that gave a disappointed litigant as many bites at the apple as he desired. Once again, in the long run, the tough procedural rules do not frustrate, but rather promote, the desired substantive outcomes.

These insights on retroactivity, form, and finality have clear implications for the current Florida electoral debacle. First, using punch-card ballots requires formulating a rule on what conduct counts as a vote. Everyone agrees that when the chad is punched out, the vote counts. But what if the chad is left hanging on two or three corners, or is dented but not removed at all? Here the first inquiry is whether the law has already chosen its rule for counting ballots. If so, then however unwise it may be, we must adhere to it religiously, to avoid the problem of conjuring up counting rules after the race has been run. The rule in Palm Beach County dates from 1990 and-in accord with national standards-unambiguously provides that a dimpled chad does not count as a vote. As a legal matter, that determination should end the matter, no matter what one thinks of its actual wisdom. The merits of this rule were extensively debated behind what John Rawls termed "the veil of ignorance"-that is, when debaters could not know which rule would help or hurt them in the future. In that setting, the only way to help themselves was to select the best and most reliable rule. Alas, once a particular dispute takes place, the candidate who is behind will always seek to expand the universe of votes, while the candidate in the lead will seek to circumscribe it. But neither has disinterested motivations, no matter how sincere his devotion to the rule of law or the will of the people. All changes in standards should be allowed only prospectively, not retroactively. As a consequence, Gore loses.

But suppose we now ask, Which rule should be adopted behind the veil of ignorance? I think Palm Beach chose the right rule on the merits in 1990 by refusing to count dimples as votes. Any electoral count requires the counting of huge numbers of ballots in a very short time. The standard punch-card ballot offers all voters an easy and clear way to express their intentions. In contrast, the dented chad is capable of two readings: The voter touched it by mistake; or the voter did not execute the intention to vote by pushing the chad all the way through. It is just too chancy to determine after the fact which scenario holds for hundreds of anonymous voters. Before the election, letting all (or some) of these votes in or keeping them all out does not bias the outcome one way or the other. But the risks of fraud and delay are both lower if these ballots just don't count, whether in Palm Beach or in the Panhandle.

In refusing to exclude the dimpled votes, the Florida supreme court cited the 1990 Illinois decision Pullen v. Milligan as allowing a manual count that reflected the voter's intention to a reasonable certainty even though the machine could not count it. But a Chicago Tribune story subsequently made it clear that even under this standard, Pullen held that dented chads did not express that clear intention. Only those chads that had been far enough removed to allow a beam of light through counted in the Illinois case-a litmus test for intention that is both simpler and more reliable than any elusive search for what the voter meant when he indented a chad. The Florida supreme court's recent decision also allowed the county boards to continue with the hand recount, noting that "our society has not yet gone so far as to place blind faith in machines" and recognizing that "humans routinely correct the errors of machines." But it is one thing to allow a recount to correct for an "error in vote tabulation," as the Florida statute requires, but quite another to use the manual recount as an occasion to change the definition of what counts as a vote. The Pullen standard deftly avoided that risk.

Nor did the Florida court pay sufficient heed to the principle of finality and the importance of res judicata. Why should we think one manual recount is less prone to error than two machine counts? Election laws generally show a high respect for the principle of finality because (from behind the veil of ignorance) most legislatures recognize that the manifest risks of bias, fraud, and mistake are magnified by any hand recount that drags on for days. Courts and administrators should resist the temptation to depart from that basic conception after battle lines have been drawn, when gloomy estimations of human conduct all look as though they are being directed personally at this or that canvassing board. It was in part to avoid these struggles that Florida's badly drafted statute contemplated a short period for recounts to take place. The same principle also explains why the Miami-Dade board made the right decision in calling off its manual recount after the Florida supreme court extended the deadline to November 26.

In truth, much of the bitterness, delay, and animosity in the Florida recount could have been avoided if the entire recount process had been conducted with an eye toward the traditional legal concerns with retroactivity, form, and finality. An actual legal dispute is the absolute worst time to reconsider first principles. If Palm Beach County had understood the prohibition against retroactive rule changes, its first decision would have been to follow its own established rule of not counting dimpled chads. That one decision would have allowed the process to come to closure without the endless disputes and controversies that have divided a nation and placed its legal system in turmoil. No one doubts that a general election is meant to ascertain the will of the voters, in close elections as much as landslides. But once this present election war subsides, all weary combatants should realize that social justice is best advanced by announcing clear and simple voting rules in advance, and then adhering to them assiduously through thick and thin, no matter what the outcome.