By design, it’s relatively difficult in the United States to register to vote compared to what exists in many other countries or compared to other systems one could imagine. One consequence of this is that incentives exist to mount “voter registration drives” that would be superfluous if we made the voter registration process something more automatic. These drives involve registering large numbers of people. And as you’ve no doubt noticed if you’ve ever done anything, it’s not possible to do anything perfectly. A process with a 95 percent success rate is a pretty solid process. But if you register 1 million voters with a 95 percent success rate, you’ll wind up submitting 50,000 forms that are, for one reason or another, bad forms. If you want to spin this out into a story of “tens of thousands of fraudulently registered voters” you can, but you’re not being very honest.
What’s needed to make a real case for voter fraud is instances of people actually voting fraudulently — people who aren’t registered to vote voting, or people voting multiple times. But year after year nobody can ever find more than a trivial number of instances of fraudulent voting. Instead, the issue is raised every year in order to raise barriers to voting by perfectly eligible voters. Raised by people who think that the high-SES voters who are disproportionately likely to overcome the barriers are likely to make better choices in the voting booth than are the low-SES voters who are disproportionately likely to be disenfranchised by broad-brush anti-”fraud” efforts.