NC-Upshot/Siena: Tie

[TheDeadFlagBlues]

As much as I approve of Nate Cohn's valiant attempts to try to build a model based on actual turnout data, rather than polls, this poll exemplifies why that is a flawed project. There's no reason to believe that the early vote shifted to Trump over the past two weeks...

[TheDeadFlagBlues]

Nate Cohn's model is a great idea but the wild changes that this poll has had on his early vote estimate suggests that the only way to implement this idea soundly is if the NYT gave him a budget that's ~10 times larger or so. I think this only works if he conducts polls every day or so with the specific intent of garnering a huge sample of early voters that's more stable. We really shouldn't be seeing fluctuations in the early vote that are this huge: these people already voted!

His revised model will probably be much closer to the final result but I'm left feeling like this will be by coincidence rather than the soundness of this whole project.