This actually isn't true, because major party candidates will be worse off, but I think he meant something like Hicks-Kaldor efficient, or positive sum.
I am also seeing several comments that try to take into account the preferences of the candidates themselves, or their parties. Why? Aren't we talking about efficiency from the perspective of the electorate? To the extent that the candidates are themselves members of the electorate, and can vote for themselves, shouldn't we ignore them in this analysis because their preferences are inherently biased?
I fully agree with you. I was just being precise in my usage of Pareto efficiency, which includes all agents in a situation. Especially because it's used so often when Hicks-Kaldor improvement is meant (which means that there's the possibility for Pareto improvement after redistribution of gains).