The Relative Value of Disenfranchised versus Fraudulent Voters

Even though I know of many other people who would be better qualified to discuss Supreme Court findings, I feel I need to say something about the math involved in today’s 6-3 decision on the Indiana voter ID law.

The case, Crawford v. Marion County Election Board, was a facial challenge of an election law passed by a Republican-controlled state legislature that would require all voters to show an up-to-date, government-issued photo ID in order to vote in-person at precincts throughout the state of Indiana. The usual conservative justices, along with Justice John Paul Stevens, rejected the plaintiff’s facial challenge because there isn’t enough evidence to support the premise that anyone is actually disenfranchised by voter ID laws. The liberal minority, minus Justice Stevens, pointed out that there also wasn’t any evidence to support the state’s premise that anyone has ever committed in-person voter fraud anywhere in Indiana in the entire history of the state. So, understanding that the evidence pickings are slim on both sides, the question is this: Is the casting of a fraudulent vote worse than the prevention of a legitimate vote?

Now I know this question assumes that individual votes actually matter, which is the case for maybe one election per decade somewhere in the U.S., but lets just say for argument’s sake that there are 1,000,000 eligible voters who would turn out for Candidate Rodham and 1,000,000 eligible voters who would turn out for Candidate Hussein, two names I just completely made up. If one person voted for Candidate Hussein twice illegally and he wasn’t caught because he didn’t have to show ID at the polls, then the election would swing to Candidate Hussein by a margin of one vote. Similarly, if one elderly voter for Candidate Rodham couldn’t get on the bus in time to get to a local government office to get a birth certificate to get a valid ID and was therefore disenfranchised, the election would swing to Candidate Hussein by the same margin of one vote. On the face of it, it would seem like a disenfranchised voter has the same value as a fraudulent voter.

One could view this like a supply and demand curve. On the left end of the chart is a situation where no election laws could exist. Vote by email, vote by Facebook profile, vote by randomly shouting in the street. Everything counts, if someone can hear you! If you wanted, you could vote dozens of times, maybe hundreds of times, depending on how much your Mafioso political system rigger was willing to pay you for your time. Total enfranchisement, but also near total corruption. On the right end of the chart is a situation where only the strictest of security measures could be enforced, and unless the government had your retinas and your dna already on file, you couldn’t vote, and even then you’d need to submit three letters of recommendation from congressmen and a self-addressed stamped envelope to the county election board before you could get your voter card. Basically, only the president could vote. No fraud at all, but also near total disenfranchisement.

So you want to be where these two curves meet. According to economics, there should be an optimum point somewhere in the middle where a law could be strict enough to prevent an unwanted amount of fraud while also preventing an unwanted amount of disenfranchisement. In evaluating whether or not the new law is too burdensome, you need to figure out where the existing law is on this curve. Unfortunately this requires data which neither side in the Indiana case was able to provide. But the scant evidence from other places on the internet (yes, supposition based on extensive internet research, as Justice Stevens would retort) reveals two nuggets of information:

1. If enacted for the entire country, this law would have prevented one fraudulent vote cast in a gubernatorial election in Washington in 2004.
2. If enacted for the entire country, 99% of eligible voters would already have met the extra burden, and 99% of the remaining balance of voters could easily meet the burden.

Therefore, assuming the population of the U.S. is 303,960,000 and 2/3rds of them are eligible to vote, the law could disenfranchise about 20,000 eligible voters (0.01%) while preventing one fraudulent ballot. Is one fraudulent ballot worth 20,000 disenfranchised voters? Is voter fraud 20,000 times as repugnant as insurmountable burdens on eligible voters?

I think we need more data.

Edit: Good to know Eric Posner must be reading my blog, except he takes the question farther, adds the variable symbol n, and, oh yeah, writes better.