While paper ballots have issues of their own, methods have been developed (scrutineering) that make traditional elections fairly difficult to steal, at least in a 1st world country such as Canada. This cannot be said for a vote conducted electronically (either on site through voting machines or over the Internet).
For example, DOS (Denial Of Service) attacks are always a worry. If candidate bent on stealing an election knew that most of "their" constituency would have exercised their franchise by a certain time, they could attempt to crash or otherwise overwhelm the computer system collating ballots after that time, thus tilting the vote count in their favor. Mind you, there are ways of getting around this: extending the voting period over several days is one such method.
But more importantly, virtual reality is, well, not really real. No matter how you might think you voted--you matter what your eyes tell you when you look that the computer screen:
...wherever the vote becomes an electron and touches a computer, that's an opportunity for a malicious actor potentially to . . . make bad things happen.
There are literally a dozen different ways that a hacker might use to tamper with vote tallies that would be undetectable to the average voter. For an account of just a few of them, read about Ed Felton's exploits with the Sequoia and Diebold systems. The only schemes that may counter the possibility of behind the screen tampering involve generating a paper record which the voter can use later to match against the online tally of their vote but, as Venezuela's Hugo Chavez recently demonstrated, that doesn't always work either.
But the news can't be all bad. Unions use e-voting to choose their leaders, and the fairly decent sized city of Markham Ontario has employed the process on a number of occasions; the 2nd time out they even got the higher turnout they were looking for.
Well, when I was doing my e-voting research back in the early 2000s, I came across a rule of thumb that might explain where and when the process can work. At the time, I was pondering a claim made by David Dill, another well-known e-voting skeptic. He argued that:
"Someone sufficiently unscrupulous, with an investment of $50,000 [U.S.], could put together a team of people who could very easily subvert all of the security mechanisms that we've heard about on these [voting] machines," he said.
...and I was discussing this claim with some of the other people involved in my research. One of them asked: "Who would risk that kind of money (and a possible jail sentence, if things didn't go as planned) for this?" This being the presidency of a fairly decent sized trade association, which involves receiving a stipend in the low five-figures, a certain amount of fame locally, and a few trips to the U.S. or overseas. The answer, of course, was: nobody in their right mind. We were dealing with business-men and, if they had that kind of money lying around, there would surely find more useful ways to invest it.
And so there is your rule of thumb: where the prize for stealing an election is is not worth the cost, e-voting can work fine! So: for union or campus elections, why not? The Mayor of Markham makes about $150,000 per year so, again, is trying to hack the vote there worth the candle?
On the other hand, a national election where the prize is control of a small (but very ferocious, mind you) army and a budget in the billions...
Well, in that case I would be worried.