But does the addition of a voting mechanism absorb proportionately more expressions of approval than disapproval?

I think so, and the evidence I was providing was an estimate of what percentage of 'negative' responses (including corrections as negative) were comments vs. downvotes, and what percentage of 'positive' responses were comments vs. upvotes.

Note that there are strong alternatives to the absorption model, since the activation energy is lower to vote than comment.