There are many reasons to move to a different system. Most of them dealing with accessibility. Electronic voting machines can present a ballot in multiple languages, electronic machines could present an audible ballot for the blind or a large print ballot for the sight impaired. Electronic machines are easier to vote on than filling in circles for those with motor skill issues.
A 2003 article presented error rates for the technologies as: 2.5% punch cards, 2.3% touch screen, 1.8% paper ballots, 1.5% optical scan and level machines. The real mission for electronic voting machines is to allow more people to vote unassisted but to do it in a way that is as accurate or more accurate as existing technology. The technology is clearly not available yet.
It seems a lot of reliability issues result from the use of touch screens and touch screen calibration. It seems that a machine with buttons around the screen (like most ATMs) would make more sense and would more closely duplicate the old lever system that proved to be so accurate. I will admit ignorance of the usability issues for this type of interface.
So the question that needs to be asked is, if a paper audit trail is so important, why is it being universally ignored? The answer lies in the reliability of the printing mechanism and the typical usage scenario that result in voting machines being idle for two years between uses. This was, I believe, the justification for leaving the printer requirement out of the first (defeated) IEEE proposed standard, but also, in some people's opinion, the primary reason why the standard proposal was defeated.
I have already voted in this election. I was offered the choice of touch screen or optical scan ballot, I chose optical scan.