When thinking about any election shaping up, one should think of two broad kinds of voters. There are “traditional” voters and “non-traditional” voters. Polls done either by polling firms or by media are usually good at figuring out what people who normally vote in elections think.
The problem comes when dealing with non-traditional voters. As a rule of thumb, you typically get 60 million Republicans, 60 million Democrats, and 100 million people sitting out and not voting. And these people have the potential to pick any candidate they want. The term The Conservative Treehouse gave to these people is “The Monster Vote”.
So when predicting elections, you have two different problems: how are the traditional voters going to vote, and how many monster voters will there be? And these problems are dealt with in two entirely different ways.
1. Inside the box – traditional voters
In 2012, the electorate was 38% Democrat, 32% Republican and 29% independent. Their party ID was 35% Democrat, 32% Republican and 33% Independent.
In 2016, the party ID had steadily over the whole year been 27.7% Republican, 30.0% Democrat and 40.6% Independent. If you assume each group will have a turnout ratio the same as in 2012, the electorate will be 32.58% Democrat, 29.56% Republican and 35.69% Independent. I go over and apply this more on the poll analysis page.
|Party||2012 ID||2012 Turnout||ID/Turnout||2016 ID||2016 projected turnout|
Now some will say “this is just like 2012 unskewing!” – but this is false. The 2012 unskewers based their prediction for the 2012 general election on the 2010 mid-terms – and Republicans always do better in mid-terms. And they predicted the electorate would be R+4.3, when in reality, as seen above, it was D+6.
I am predicting the electorate will be D+3. Moreover, my adjustments, as of September 15, only give Trump 1.32 points compared to other polls.
In addition, I have compiled and adjusted battleground state polls with the D+3 national sample assumption, for example making Pennsylvania a D+5 state and New Hampshire an R+3 state (yes, that’s real and NH has been R+3 many times in the past as well), and based on that, this is my belief of what would happen if the election were held TODAY and only traditional voters were allowed to vote:
Prediction of how “traditional voters” would vote if the election were held today
Trump would need to win New Hampshire or Virginia, Hillary would have to win both.
One of the reasons that I am slightly biased toward Trump in this map is due to the enthusiasm gap. Obviously it wasn’t enough to get Romney the win in 2012 (since Romney was battling against Obama’s monster vote), but in my mind I treat it as half a point.
Now if the polls keep trending this way, Trump can win simply with inside-the-box voters and doesn’t have to rely on non-traditional voters one bit.
2. Outside the box – non-traditional or “monster” voters
There is evidence that Donald Trump is benefiting from non-traditional voters in a yuge way. The only way I know of to infer the existence of these voters for the general election is primary turnout.
Unfortunately, the oracle known as FiveThirtyEight has planted the seed that primary turnout doesn’t matter. They claim this with a very simplistic, binary analysis of whether or not the party with a bigger increase in turnout and/or more absolute turnout won the election.
Obviously what you would want to do is look at the correlation – to see the degree to which primary turnout predicts general election turnout. And you would do it state-by-state, since that’s how presidential elections actually happen.
And our very own John Macintyre showed quite thoroughly that there was a .542 correlation between GOP primary turnout and general election turnout, and a .520 correlation for the DNC. This means that for every 1,000 new primary voters, we should expect 542 more general election voters for the GOP and 520 for the DNC.
As it happens, the primary effect may not have been enough to swing any election up to this point, but that doesn’t mean it didn’t make some elections closer than they otherwise would have been, and it doesn’t mean that it won’t put Trump over the top this November.
To dismiss the primary effect would be like writing off a football team that lost to Alabama, Ohio State and Oregon by 3 points each time as a team that “doesn’t win”.
Overall, in 2016 the Republicans had a 45.02% increase in primary over 2008, while the Democrats had an 18.92% decline. And, based on the 2008 election, we can use JM’s model to predict general election turnout.
|State||Republican Turnout based on 2008-2016||Republican Turnout based on 2012-2016||Democrat Turnout based on 2008-2016||Obama 2012 turnout|
|Wisconsin||1.584 million||1.526 million||1.622 million||1.621 million|
|Pennsylvania||3.071 million||3.107 million||3.019 million||2.990 million|
|Ohio||3.176 million||3.095||2.371 million||2.828 million|
|Virginia||2.016 million||2.234 million||1.854 million||1.972 million|
|North Carolina||2.028 million||2.366 million||1.915 million||2.178 million|
|Florida||4.270 million||4.535 million||4.261 million||4.238 million|
For Republicans, the changes are from both the 2008 and 2012 primaries as they were both competitive. For Democrats we can only use 2008.
Now don’t take these numbers too literally. The fact that one party should do better based on this primary turnout model is just one factor to consider. And I am suspicious of numbers that predict Hillary 2016 doing better than Obama 2012. Models aren’t magical oracles, just indicators. That said JM did a fantastic job with his model.
I believe there is going to be some sort of “monster vote effect” – but if it is as real as this, this election is over. We are in uncharted territory with the scale of the primary turnout changes from 2008 to 2016, and there is probably a logarithmic effect for increased primary turnout – for no other reason than that any increase in primary voters reduces the number of non-primary voters who could add to the total in the general election.
In 2008 Obama wasn’t on the Michigan primary ballot, which makes the 2008 Michigan primary meaningless, and so we can’t do this with Michigan. Colorado has caucuses, so we can’t make any predictions about Colorado either.
After considering the “monster vote”, this is how I think the election would go if held today:
Election prediction for today given the primary turnout effect
And so I find myself predicting a very large Trump win if the election were held today.
That said – this election is far from over. Julian Assange hasn’t dropped his October surprise yet, and Trump will probably improve his numbers with the debates (not from being so much better than Hillary, I don’t think he even needs to “win” to improve his numbers from it, and Hillary is a lot better on her feet than people give her credit for. But he will gain by simply being a normal person and undoing the impression that he is a maniac).
And thus Trump will probably gain more ground in the next 50 days. Remember back in the desperate days of early August, Trumpsters were banking on Assange’s revelations and the debates. Trump closed the gap before and without those things – but those things are still going to happen. And I am now nervous because I feel herded into a dangerously optimistic outlook, which I feel is fragile and can at any moment be taken away.
Based on all of this, I find myself pulled inexorably toward a prediction I never thought I would make: Trump landslide.