karma balances often get downvoted less than the vote options get upvoted, so this will tend to over-estimate how many people no longer dissent.

To me, when a karma balance is downvoted less than poll options are upvoted, it means that people think running the poll deserves some karma. This does not overestimate people who have reacted to voting patterns, since that number does not come from the karma balance. If someone (who has NOT reacted to voting patterns) wants to give karma for running the poll, they would upvote the karma balance, not the voting comment

Also, the purpose of the poll is to see whether a relatively high or relatively low amount of people have reacted to the voting patterns this way. Exact numbers are not needed.

I have a proposal for a new structure for poll options:

The top-level post is just a statement of the idea, and voting has nothing to do with the poll. This can be omitted if the poll is an article.

A reply to this post is a "positive karma balance" - it should get no downvotes, and its score should be equal to the number of participants in the poll.

Two replies to the "positive karma balance" post, you downvote one to select this option in the poll.

This way voting either way in the poll has the same cost (one downvote), the enclosing post will have a high score (keeping it from being lost), and the only way to "corrupt" the poll results without leaving a trace [downvote the count post and upvote one of the option posts] simply cancels someone's vote without allowing you to make your own.