One of the biggest myths (and there are very many) surrounding my channel is that I do not use "data" in my predictions. This could not be further from the truth. In fact, it is usually my critics who are not using enough data, usually by relying on polls too much. Therefore, in this article I will look at all of the methods that people should be looking at in terms of gauging the true barometer of this election.
First of all, yes, polls are important to look at. They are not always the most accurate data points in July of an election cycle, however, and nobody at this point should be basing their predictions solely based on polls. Hell, even in late October, several swing states had Hillary Clinton up by 8 points or more that eventually went to Trump. There is plenty of time between now and the election. We have not even had either convention, or any of the three debates. I specifically like to say that the polls do not matter until 10 days after the second convention. And even then, they don't matter that much until the week after the third debate. Late October is a turning point where pollsters screen more for likely voters (the past few cycles, these have leaned Republican). Polls also currently are not always weighted properly by Party ID, and often have contradicting crosstabs surrounding ethnicity and support among education demographics. That is why it is very important to weigh them when we can and take them with a grain of salt.
Second of all, we must analyze crosstabs within polls that could determine the way the race can move. These include things such as "soft support," "hard support," or enthusiasm. For example, Biden's support base has not made their mind up as much as Trump's base has, and Biden's enthusiasm numbers are much lower than Trump's across nearly every single poll. Biden's "strongly favorables" are even lower than Hillary's were in 2016. Trump also is the one most people trust with an economic recovery, and Biden is who people trust with their health care. These are consistencies that need to be analyzed and interpreted which can determine the way the race moves.
Third of all, we need to analyze party registration data and long-term trends. The fact that Republicans are doing very well in places like Florida, Pennsylvania, and North Carolina since 2016 proves that some of the polls from these states may be missing the grassroots action on the ground. Democrats have improved in places like Arizona and Georgia, albeit marginally. The fact remains that the rust belt is moving right, and the sun belt (with the exception of Florida and North Carolina, at least for now, both states which have plenty of rust belt transplants) is moving left. This will be reflected no matter what in how the states vote relative to the national popular vote, with few exceptions.
Fourth of all, we need to look at primary turnout. Primary turnout increases nearly every cycle, so it is no surprise that Joe Biden will receive nearly 18 million votes despite being one of the weakest, most boring individuals that a party has nominated in recent history. Nevertheless, Donald Trump has received nearly 18 million votes himself despite several populous states cancelling their primaries and binding their delegates to him. Obama in 2012 received just 6 million votes, and Bush in 2004 received 8 million votes. Both cruised to re-election. Donald Trump has brought immense energy to the primary turnout game, and that cannot go unnoticed. Primary turnout is not everything, but it is not "nothing" either, just like any other method of predicting elections.
Fifth of all, looking at historical precedents and election models are not definitive, but still worth analyzing. Models like Moody's Analytics (which predicted a Trump landslide, but predicted a Clinton landslide in 2016), Lichtman's Keys Model (which previously predicted a Trump Victory in 2016, but is unclear for 2020), and Helmut Northport's Primary Model (which was right in 2016 and predicts a Trump victory, yet likely overestimated his magnitude in both cycles) are all good resources. Also, looking at the previous incumbent trends, it proves that by precedence the incumbent party has a greater chance of winning after one term.
The bottom line is that using data is important, but polling data alone is not the only thing that should be used to predict an election. And even so, the polls must be properly analyzed on many levels while using them. This election will more likely than not go down the wire. I got 2016 right (except for two states, Michigan and New Hampshire, which ended up being the closest). My 2018 predictions were better than most. And after the disaster known as Super Tuesday, I nailed nearly all of the predictions for the next two weeks down to the decimal points. It's funny that the people who got them wrong are the people calling me an idiot. Time will tell.