We are not doomed: Why the polls, and the polling aggregators, are wrong

The polls are wrong. The people who Trump have offended are more likely to vote than the polling models predict. The amount of people who voted is not included accurately in the polls, and these people have mostly voted Democrat (by demographics). It's going to be a Clinton landslide.

There have been reports on publicly conducted polls on a near hourly basis over the past two weeks. These are likely to dry up today and tomorrow, as polling these polls are mostly paid for by media outlets, and they tend to be less newsworthy and attention grabbing after people have voted.

I prefer fivethirtyeight.com, and Nate Silver, for getting my polling information. They take publicly available polls, and (with a pretty transparent methodology), combine them to form a kind of "poll of polls". They also have a prediction engine that combines market conditions, presidential favourability, incumbent parties, etc.

The narrative fed to us by most media organizations is that the race is tightening, and that Trump is more or less as likely to win as Clinton. This is essentially the same argument many media companies make for human caused climate change: one side has vastly more science and expertise behind it than the other, but both are presented equally.

Here are two factors that polling companies likely have wrong, because they are essentially impossible to incorporate, that indicate a Clinton blowout is coming on Tuesday:

1. Public polls can't accurately account for people who have already voted. As of Friday, there are reports that anywhere from 22-40 million people have already voted. Most key states (Florida, Ohio, North Carolina, Nevada, Texas, Arizona, etc.) report that the demographics largely favour Clinton compared to voter turnout in 2012. 
2. Demographics. Each poll, and each aggreggator of polls, uses a model of what they think the electorate will look like in 2016. They can use whatever information they want in creating these models, and it determines what identities are included in their reporting. This changes the number of people who are actually included in the polls:
1. Women
2. Men
3. Hispanic/Black/Asian/White
4. College Educated/Not College Educated.

If you change the way any of these demographics are included in a poll, it wildly changes the result. Include 5% more hispanics, 5% less white voters, and all of a sudden Trump looks like he's in big trouble. In this cycle, with how offensive Trump has been, it's impossible to predict how the electorate will be different in 2016 compared to 2012.