2016 Postmortem

All poll data (including exit polls) is normalized against the modeled turnout for the election. This risk happens every time you take a sample and attempt to project it to a larger population. The same pollsters ran the exit polls and the pre-election polls. They almost certainly used the same models of the expected turnout.

The models assumed that turnout would be at least somewhat similar to 2008 and 2012. We now know that was wrong.

Interestingly enough, the poll that was most accurate in 2012 (The LA Times tracking poll) was the most accurate in 2016 as well; and has a distinctly different methodology.

The imagined parallel between the DNC and Podesta hacking and supposed voting machine hacking is absurd. Email systems are always connected to the internet and have a high volume of traffic. Voting machines are never connected to the internet, most have no external network capability at all. Some (like the ELECTronic machines used in PA) can't take software updates at all, even in person. To hack the voting machines, it would have taken a team of hundreds of operatives on the ground physically forcing access to tens of thousands of voting machines. Its just so very improbable.