The most likely way for Hillary Clinton to not win the presidency may be a tie between Secretary Clinton and Donald Trump. This is because, when one looks at the data a number of ways, and makes various adjustments, Clinton wins, often just barely, most of the time, except in what appears to be the worst case scenario. That scenario is Clinton losing most of what are called “Battleground States” — but for the most part, only those that are truly in contention, so it is quite possible — but retaining her “firewall” states, the states she really can not possibly lose. That puts Clinton 3 points ahead of the 270 required to win. But then, in this scenario, the most likely bluish state to switch to red, New Hampshire, goes for Trump. When that happens, the Electoral College Vote becomes 269-269, and the Electoral College becomes the Electoral College Prank.
What happens then? The House attempts to decide who will win. If that happens, each state gets one vote (or zero, if they can’t decide). Even if the Democrats win the house back from the Republicans this election, Republicans will theoretically decide the outcome, because Democrats are concentrated in the more popular states. On a state-by-state basis, most states are Republican.
That does not mean that the Republicans will vote for Trump automatically. They have to chose among the top three Electoral candidates (while the Senate, meanwhile, choses among the top TWO VP candidates). Who knows what will happen?
You might think this is unlikely. Until I did my analysis this morning, I thought it was possible, but unlikely. I now realize that the chances of an electoral tie are pretty darn good. (And by pretty darn good, I do mean probably less than one in ten, but that’s still pretty darn good for something that has only happened once before.)
Let’s look at all the numbers.
As you know I have a model. I mentioned weeks ago that near the end of the election season, my model would converge on the polls, because it is calibrated to the polls, but only uses the better and more recent polling data. Today, I decided to use the final adjusted polling estimate provided by FiveThirtyEight, because, a) they are good at adjusting and evaluating polling data, and b) there is now enough information to use polling data from pretty much any state. Still, there are some weak states, and there are other uncertainties, so feeding polling data into my model provides a semi-independent look across the states (it is quite possible for the polls to put a state in one column but my model to reverse that).
(Note: my model does not use polling data from Utah or Hawaii, because those states are too different from all the other states.)
So, here I’m going to use two separate sets of results, polls and my model. My model’s multiple R-squared value is really high (0.9838) and the polling results and model results are almost identical, but not quite. Given the strength of my model during the primaries, I trust it more than the polling data. Also, my model foretold many things that the polls finally caught up with, over the last several days, such as the weakness of North Carolina as a Clinton state. Well, not many things, but that one thing and maybe a few other things.
This is what the current polls say about Clinton’s chances in the race. If we take all the polls, and assign every state where Clinton beats Trump to Clinton, we get this:
As noted on the map (made using 270 to win’s excellent tool), Clinton, according to the best available analysis of current polls, would win by only 3 electoral votes. I’ve seen this coming for some time, and despite lots of arm waving saying it is not true, this is the most current, scientific, likely most accurate estimate.
The weakest state among the blue states on this map is New Hampshire. Look closely at New Hampshire on election night. If this map is shaping up as indicated here, AND New Hampshire looks weak, like maybe a Trump win, then we may well have the ultimate election night hangover on Wednesday. An electoral tie.
All the nay sayers out there (you know who you are) who have been telling me that my model must be wrong, because the polls show Clinton doing much better than my model, etc. etc., take heed now. That map, above, was from your precious polls. The following map is from my model, and it has a somewhat more secure win for Hillary Clinton.
I’m giving Florida and Nevada to Clinton, and New Hampshire is more secure. Frankly I think the most likely scenario is either one of the above two maps, or something in between, and that’s pretty much what is going to happen on election night. A trivial and incorrect way to calculate the likelihood of a tie is to look at all the different combinations (moving NH, NV, and FL around) but that is dumb, so I’m not going to do it. The extremes are probably less likely than the other combinations.
One prediction comes out of this that is rock solid. Tuesday night and Wednesday morning are going to be nail biters.
But wait, there’s more. Let’s have another look at the map, but applying the uncertainty in my model, in order to get one possible Election Night Bingo Card version. This map shows what states to watch, because they are the ones right in the middle between the two candidates.
By the way, recent information out of Florida seemed very very positive with respect to that state. But that is only one study, using a methodology and a set of data never before used, in a highly dynamic and changing system, in an untrustworthy state. Comment such as “Yeah, but Florida is in the bag for Clinton” will be frowned upon.
Here’s the same deal, but based on polls instead of my model:
Now, lets try some Magical Thinking. From Trump’s perspective, consider that the polls have been shifting by about one percentage point towards Trump or away from Clinton per week over the last few weeks. So, let’s move one percentage point from Clinton to Trump across all the polls and see what we get.
We get this, the Map from Hell, in which Trump does not win, but the rest of us lose anyway.
The second Magical Thinking scenario involves the idea that Clinton, and the Democrats have a real ground game going, and Trump does not. In this scenario, we move 2.5% from Trump to Clinton across the board to reflect this political reality. This may be the case, but it could also be, as noted, wishful magical thinking. And, it looks like this:
A lot of people have been talking about a Clinton Landslide, but this is the best you are likely to get. And, if you want to call this a landslide, feel free, but it isn’t and you would be wrong.
And, finally, your election night watch list. This map shows as blue every state that remained blue in all of the above analyses, and as red every state that remained red in all of the above analyses. The unknown state are, therefore, states that have either moved back and forth depending on how you look at the data, or what are within a short distance, either by polling or by my model, of those states. This is actually a pretty robust list. I don’t expect any state not brown on this map to move, and some of the brown ones won’t either (Colorado will be Clinton, Georgia will be Trump). But, if things are wonkier and wackier than our imaginations even now let us allow, who knows…
from ScienceBlogs http://ift.tt/2eEkFpm
The most likely way for Hillary Clinton to not win the presidency may be a tie between Secretary Clinton and Donald Trump. This is because, when one looks at the data a number of ways, and makes various adjustments, Clinton wins, often just barely, most of the time, except in what appears to be the worst case scenario. That scenario is Clinton losing most of what are called “Battleground States” — but for the most part, only those that are truly in contention, so it is quite possible — but retaining her “firewall” states, the states she really can not possibly lose. That puts Clinton 3 points ahead of the 270 required to win. But then, in this scenario, the most likely bluish state to switch to red, New Hampshire, goes for Trump. When that happens, the Electoral College Vote becomes 269-269, and the Electoral College becomes the Electoral College Prank.
What happens then? The House attempts to decide who will win. If that happens, each state gets one vote (or zero, if they can’t decide). Even if the Democrats win the house back from the Republicans this election, Republicans will theoretically decide the outcome, because Democrats are concentrated in the more popular states. On a state-by-state basis, most states are Republican.
That does not mean that the Republicans will vote for Trump automatically. They have to chose among the top three Electoral candidates (while the Senate, meanwhile, choses among the top TWO VP candidates). Who knows what will happen?
You might think this is unlikely. Until I did my analysis this morning, I thought it was possible, but unlikely. I now realize that the chances of an electoral tie are pretty darn good. (And by pretty darn good, I do mean probably less than one in ten, but that’s still pretty darn good for something that has only happened once before.)
Let’s look at all the numbers.
As you know I have a model. I mentioned weeks ago that near the end of the election season, my model would converge on the polls, because it is calibrated to the polls, but only uses the better and more recent polling data. Today, I decided to use the final adjusted polling estimate provided by FiveThirtyEight, because, a) they are good at adjusting and evaluating polling data, and b) there is now enough information to use polling data from pretty much any state. Still, there are some weak states, and there are other uncertainties, so feeding polling data into my model provides a semi-independent look across the states (it is quite possible for the polls to put a state in one column but my model to reverse that).
(Note: my model does not use polling data from Utah or Hawaii, because those states are too different from all the other states.)
So, here I’m going to use two separate sets of results, polls and my model. My model’s multiple R-squared value is really high (0.9838) and the polling results and model results are almost identical, but not quite. Given the strength of my model during the primaries, I trust it more than the polling data. Also, my model foretold many things that the polls finally caught up with, over the last several days, such as the weakness of North Carolina as a Clinton state. Well, not many things, but that one thing and maybe a few other things.
This is what the current polls say about Clinton’s chances in the race. If we take all the polls, and assign every state where Clinton beats Trump to Clinton, we get this:
As noted on the map (made using 270 to win’s excellent tool), Clinton, according to the best available analysis of current polls, would win by only 3 electoral votes. I’ve seen this coming for some time, and despite lots of arm waving saying it is not true, this is the most current, scientific, likely most accurate estimate.
The weakest state among the blue states on this map is New Hampshire. Look closely at New Hampshire on election night. If this map is shaping up as indicated here, AND New Hampshire looks weak, like maybe a Trump win, then we may well have the ultimate election night hangover on Wednesday. An electoral tie.
All the nay sayers out there (you know who you are) who have been telling me that my model must be wrong, because the polls show Clinton doing much better than my model, etc. etc., take heed now. That map, above, was from your precious polls. The following map is from my model, and it has a somewhat more secure win for Hillary Clinton.
I’m giving Florida and Nevada to Clinton, and New Hampshire is more secure. Frankly I think the most likely scenario is either one of the above two maps, or something in between, and that’s pretty much what is going to happen on election night. A trivial and incorrect way to calculate the likelihood of a tie is to look at all the different combinations (moving NH, NV, and FL around) but that is dumb, so I’m not going to do it. The extremes are probably less likely than the other combinations.
One prediction comes out of this that is rock solid. Tuesday night and Wednesday morning are going to be nail biters.
But wait, there’s more. Let’s have another look at the map, but applying the uncertainty in my model, in order to get one possible Election Night Bingo Card version. This map shows what states to watch, because they are the ones right in the middle between the two candidates.
By the way, recent information out of Florida seemed very very positive with respect to that state. But that is only one study, using a methodology and a set of data never before used, in a highly dynamic and changing system, in an untrustworthy state. Comment such as “Yeah, but Florida is in the bag for Clinton” will be frowned upon.
Here’s the same deal, but based on polls instead of my model:
Now, lets try some Magical Thinking. From Trump’s perspective, consider that the polls have been shifting by about one percentage point towards Trump or away from Clinton per week over the last few weeks. So, let’s move one percentage point from Clinton to Trump across all the polls and see what we get.
We get this, the Map from Hell, in which Trump does not win, but the rest of us lose anyway.
The second Magical Thinking scenario involves the idea that Clinton, and the Democrats have a real ground game going, and Trump does not. In this scenario, we move 2.5% from Trump to Clinton across the board to reflect this political reality. This may be the case, but it could also be, as noted, wishful magical thinking. And, it looks like this:
A lot of people have been talking about a Clinton Landslide, but this is the best you are likely to get. And, if you want to call this a landslide, feel free, but it isn’t and you would be wrong.
And, finally, your election night watch list. This map shows as blue every state that remained blue in all of the above analyses, and as red every state that remained red in all of the above analyses. The unknown state are, therefore, states that have either moved back and forth depending on how you look at the data, or what are within a short distance, either by polling or by my model, of those states. This is actually a pretty robust list. I don’t expect any state not brown on this map to move, and some of the brown ones won’t either (Colorado will be Clinton, Georgia will be Trump). But, if things are wonkier and wackier than our imaginations even now let us allow, who knows…
from ScienceBlogs http://ift.tt/2eEkFpm
Aucun commentaire:
Enregistrer un commentaire