The data was opt-in (redditors had to specifically elect to enter this dataset), so it's probably biased somewhat.
I don't think that's a good assumption. I surely have more than 2000 votes on reddit, but some of them are on political stories and I don't particularly want a public record of that.