In the wake of closely contested elections, calls for laws that require voters to present photo identification as a condition to cast a ballot have become pervasive. Advocates tend to rely on two rhetorical devices: (1) anecdotes about a couple of elections tainted by voter fraud; and (2) “common sense” arguments that voters should produce photo identification because identification is required to board airplanes, buy alcohol, and engage in other activities. This Article explains the analytical shortcomings of anecdote, analogy, and intuition, and applies a cost-benefit approach generally overlooked in election law scholarship. Rather than rushing to impose a photo-identification requirement for voting, policymakers should instead examine empirical data to weigh the costs and benefits of such a requirement. Existing data suggest that the number of legitimate voters who would fail to bring photo identification to the polls is several times higher than the number of fraudulent voters, and that a photo-identification requirement would produce political outcomes that are less reflective of the electorate as a whole. Policymakers should await better empirical studies before imposing potentially antidemocratic measures. Judges, in turn, should demand statistical data to ensure that voter identification procedures are appropriately tailored to deter fraudulent voters rather than legitimate ones and do not disproportionately exclude protected classes of voters.