All the justices agree that states must "balance" the benefits and costs of voter ID requirements. The benefit is reduction in voter fraud; the cost is disenfranchisement of those who cannot obtain the ID. No one explains what an appropriate balance is, and the answer is far from clear.
To see the problem, suppose that a state can choose an ID requirement along a sliding scale from zero to maximal, where zero means something like "swear that you are eligible to vote and this is the only time you are voting" and maximal means something like supplying a birth certificate and a passport and a driver's license and thumbprints and dental records and retinal scans, etc. As the state increases the strictness of the ID requirement, the risk of fraud declines but the degree of disenfranchisement increases. The question is, what is the constitutionally permissible range on this scale?
It is tempting to think that the social cost of a single fraudulent vote is equal to the social cost of a single disenfranchised voter, and so the rule should be set at the point beyond which more voters would be disenfranchised than fraudulent votes prevented. But this conclusion would be hasty. We might think that even a little fraud throws the whole system in disrepute; or we might instead worry that disenfranchising people is much worse than tolerating a little fraud.
There are more complications. Suppose you think that the majority should rule, and so if a community has 100 eligible voters, then the candidate preferred by 51 should prevail. A strict ID rule, by excluding the poor, would bias results in favor of candidates supported by the rich. But what would be the effect of an insufficiently strict rule? Would the fraudsters favor the candidate of the rich or the candidate of the poor? If you are tempted to say that fraudsters would be paid by supporters of the rich candidate, then you should have second thoughts about supporting a weak rule. A weak rule would, along this dimension, hurt the poor by making it easy for the rich to finance voter fraud.
The current debate assumes that the majority party will choose a rule that keeps the minority party out of power. But once we agree that the majority party can enact anti-fraud rules, we again need to ask how to determine when the majority goes too far. Suppose that the majority belong to the Republican party, and that an incremental tightening of the rule eliminates 10 Democratic voters but also eliminates n cases of fraud. For the new rule to be permissible, can n be 10, or does it have to be 100 or 1000? Could n be 1? Does it matter who the fraudulent voters would have voted for? If they would have voted for Republicans, and n > 10, then the move injures Republicans. Perhaps the safest assumption is that fraud would favor neither party. But its cost would still need to be quantified; otherwise, it is impossible to choose a reasonable n . And how do we quantify the cost of fraud that, because it favors neither party, has no impact on the election?
Many people say that fraud undermines voters' confidence in the system. But so what? Do voters with less confidence vote less? If both the rich and the poor, or Republicans and Democrats, vote less, then reduced confidence won't affect the outcomes of races. And how do we decide how much fraud has to occur before confidence is seriously weakened?
The fact is that no one agrees on what voting systems are supposed to do, nor does anyone really understand how much voting fraud is tolerable and how much is too much; so no one can agree on what the costs and benefits of anti-fraud rules are. Perhaps this is the real reason that the Court is signaling that it wants to have nothing to do with this problem.