Happy to clarify, @NateSilver538. 1/ You claim we show people “weight” probabilities toward 50/50. Fine. But in *every condition,* including vote share only, people report a likelihood of victory close to 50/50. They *expect* a candidate’s odds to be close to 50-50.
2/ Here’s the plot: 
3/ When we tell them "your candidate has an 87% chance of winning,” it sounds *so far off from prior expectations* that we see large effects across all outcomes. Because most people don’t understand probability in the way that forecasters define it, as this and other work shows.
4/ But here’s the thing—that’s one of our more trivial findings. Far more important is this: win-probabilities make people *substantially more certain* about the outcome than vote share projections, especially when their candidate is ahead. (These are 95%CIs, not SEs).
5/ And critically, the closer you are to forecasting a blowout, the less likely people are to vote. But you *don’t see that relationship* with vote share projections.
6/ In 2016, win-forecasts may have disproportionately affected democrats - 1) exposure favored the left & 2) effects from study 1 are stronger for candidate who was ahead. And Democrats were particularly likely in 2016 to say that Clinton would “win by quite a bit.”
7/ In your podcast, you urge reviewers reject our paper. That’s not the way it works—you carefully read, considering the merits, & give feedback. It’s not about what others say about a paper. And anyway telling an academic what to do is like trying make water run uphill.
8/ BTW here's a link to the paper, w/ amazing co-authors @seanjwestwood and @ylelkes : papers.ssrn.com/sol3/papers.cf…
