Thread: 1/ Election predictions have consequences—for voters AND officials. With @Comey’s book we see that confidence in a @HillaryClinton victory may have influenced decision to reveal FBI investigation into emails in final days of campaign. H/T @ylelkes, @gdebenedetti 
2/ Not the first time we’ve seen effects like this — @RyanDEnos & @eitanhersh have shown that other elites—campaign staff—make strategic decisions based on whether they perceive the race to be competitive, and if so better represent their constituents. cambridge.org/core/journals/…
3/ As we (@ylelkes, @seanjwestwood) show in a recent paper, what was different in 2016 was the massive amount of coverage devoted to election forecasters (@googlenews data). Paper here: papers.ssrn.com/sol3/papers.cf…
4/ And the win-forecasts they produce *dramatically* increase the certainty that the leading candidate will win
5/ They also increase the perceived margin of victory.
6/ Folks these are *big* effects based on a national probability sample. Odds we found signal in noise (p-values) lower than machine precision for many relevant tests. Those are 95% CIs in the plots, not standard errors. And they don't change w/ education level.
7/ Theres a lot of research showing how perceived electoral closeness boosts turnout and how people more likely to stay home when they perceive a blowout
8/ And we definitely find that blowout forecasts keep people from voting, in a behavioral economic experiment simulating an election.
9/ Effect on officials is something we've suspected, and the evidence, even if anecdotal, is now growing. Hoping for more research into this.
