Forecasts didn't cost Hillary Clinton votes among overconfident Democrats, but lowered turnout among Republicans confident their candidate would lose & stimulated turnout among Democrats confident their candidate would win

Revisiting the Link between Expected Election Outcomes and Turnout. Elizabeth C. Connors, Jacob A. Martin, John Barry Ryan. http://www.elizabethchaseconnors.com/images/Election%20Outcomes.pdf

Abstract: Westwood et al. (2020) causally demonstrate that probabilistic forecasts reduce beliefs about electoral competition, which, in turn, results in a lower propensity to vote. They argue this may have cost Hillary Clinton votes among overconfident supporters. Using the American National Election Study (ANES) from 2004 to 2016, we find that believing an election will not be close can affect turnout. The strongest evidence, however, suggests only those who believe their party’s candidate will lose by quite a bit are less likely to vote. Analysis of validated vote in 2016 does not support the conclusion that probabilistic forecasts cost Hillary Clinton votes among overconfident Democrats. The results suggest, if the probabilistic forecasts had any effect, they either lowered turnout among Republicans confident their candidate will lose or stimulated turnout among Democrats confident their candidate will win.