I recently wrote a piece that argued there is an anti-winner bias in polling. For races expected to be competitive, I noted that the predicted poll margin underestimated the actual vote share for the winner.
I’ve thought more about this issue and talked with some people about it (including the Monkey Cage’s own Andrew Gelman), and I don’t think the claim holds up to closer scrutiny. It may even be a statistical illusion. CONT.
Eric McGhee, The Monkey Cage