I have to take back one of my criticisms. If there were a way of ensuring that a person's vote wouldn't be manipulated or lost (like with Ronald L. Rivest's methods ThreeBallot, VAV, and Twin), and adding fictitious votes could easily be detected, then a national popular vote might be resistent to fraud. Throw in a few observers from neighboring states -- who would have an interest in making sure everything was aboveboard -- and it would be better than the paperless electronic methods.