Above is my latest electoral college projection.
This uses the technique previously described. However, instead of using RCP averages for all polled states and then using extreme (non-tossup) states to develop the regression model, this method uses only polling from states with one or more recent poll, and only with good polls. these poll numbers are then "predicted" by black/hispanic/white/Voted_Romney numbers, and that generates a model, based on just over 20 states, designed to predict all the states.
As expected, the r-squared value is much lower using this method, but this method does not violate any important statistical laws like the last one did.
Most of the polling data pre-dates the revelation of Trump's interest in sexual assault, last Friday, and of course, Monday's "I'll throw my opponent in prison when I win" debate on Sunday. If you believe those events influence the election further, then you can figure this is a conservative estimate from the perspective of Clinton.
All of the blue states, both shades, are projected to go to Clinton, but I left the three closest to 50-50 in light blue.
I suspect the most controversial state here is actually Iowa, which seems to be throwing some sort of hissyfit in the polls.
And this, of course, is why my model is different from everyone else's. The polls are used in this case to calibrate (in the absence of earlier results, like could be done in the primary!) but the actual prediction then does not use the polls directly. So, even though a recent poll showing Iowa as Trump, the model does not, because the model does not lie like the Iowans do, apparently!
- Log in to post comments
You are definitely going out on a limb here. I could see Georgia going blue--in 2012, IIRC, it had the second smallest margin (after NC) of states Romney carried. Last week that would have been a long shot, though I expect the chances of that have improved significantly. However, I don't see SC and MS going blue while AZ stays red. Not unless likely voter models are both underestimating black turnout and overestimating Hispanic turnout by significant margins. The former might be true; I doubt both of these things are true. Remember that polling from before the Access Hollywood video (let alone Sunday's debate) was indicating that Arizona is in play this year.
The main problem is that you have no way to validate your model until the actual election results come in. People like Sam Wang and Nate Silver have a track record, so they have some data points already in hand to verify their models. (Although I think Silver's error bars are too big--his predictions have been rather too sensitive to news noise.) Wang got 49.5 out of 50 states right in 2012 (he declared Florida too close to call; it went narrowly for Obama).
I hope you are right about Iowa staying blue, but you are bucking the trend on that one, too.
Actually, what I'd really like to see is Hillary win the electoral college 537-1 (I'll give Trump NE-03, to preserve the tradition that only George Washington gets a unanimous EC vote), but barring a complete GOP collapse, that isn't going to happen.
I'd like to see more evidence that I'm bucking a trend in Iowa. As far as I can tell I'm bucking a poll, and that poll is way skewed by a high third party effect, which often goes away. I'm not worried about Iowa.
Arizona as well as Georgia and some of the other states are in the middle, all within 2% either one side or the other of 50-50.
As was the case with the primary, I suspect I'll be wrong on a few states but very close in the actual percentage differences, so even if I'm flipped backwards on the electoral college votes for a couple of states, I'll still have the percent distribution within a point or two. (Or I'll be totally wrong)
Other models, such as those you mention, have history and validation, and that history and validation made them often wrong in the primaries. My model does not need no stinking validation because although I use some statistical tools, it is NOT a statistical model. It is little more than a (somewhat) simple arithmetic construct.
Next iteration I'll do more to indicate which states are in the middle zone, and/or include the list of the actual percentage data. I want a bit more refinement first.
LOLOL. Fun read. Especially the comments.
That would be particularly useful. Especially if you can estimate what your error bars might be. What you predict to be a Clinton +2 state could well go to Trump, or a Trump +2 state to Clinton.
What seems to be happening in Iowa is that the state has a relatively large fraction of old white people without college degrees. That's Trump's core demographic. There are similar issues for Democrats with Ohio and ME-02, which have only turned blue on prognosticators' maps as Hillary has opened up a national lead. But in Iowa there is a reasonably popular Republican governor (Branstad) who openly supports Trump. That's not true of Ohio (Kasich is reasonably popular but is one of the few openly anti-Trump Republicans) or Maine (LePage openly supports Trump but is not very popular).
Since I'm calibrating to polls, but then using a deterministic model, one could guess that the error bar is about 5 points. However, using the behavior of this kind of analysis from the primaries, my range of accuracy was closer to one point, so a 2 SD error bar might be comfortable about plus or minus 2 points.
I wonder if Branstad will react to the latest Trump news in some wa or another.
Note: Today's polls, from a handful of states, are in close agreement with the predicted values from the model, so nothing changes yet.
I shared this with a few of my classes for reading and discussion: I thought you might be interested in it as well (and if you've already seen it, my apologies for clogging up the site.)
Dean, if that (statistical improprieties) becomes "How one 19-year-old Illinois man is distorting national election results" (legal improprieties), I'm going to start worrying...
(You're not clogging up the site, but I think we can posit a few notorious others who have been doing so.)
I would be very interested to hear how the people running that poll feel about this thing: they try to do everything right, and make all of their work transparent, and despite all of their good intentions...
Sure, I got there-here PID loop... Let's compensate for a weak signal by just turning P "up to 11"...
What could possibly go wrong?
You've got the coal states (except Illinois) going red, which makes sense to me. Perhaps Illinois doesn't mine that much coal any more.
It is truly amazing how wrong this map is ?. I will make a personal wager with anyone that Trump will take Ohio , Georgia , Nevada and many more the crazy liberal who made this map gave to Hillary.