A per-state prediction methodology (Bush in bad shape in FL)

[Shira]

A per-state prediction becomes irrelevant if there is a clear nationwide victory of one of the parties by at least 53:47 ratio. In this case the winner will take ALL of the tossup states.

If, on the other hand, the election outcomes would be balanced like in 1960, 1976 and 2000, then the numbers would perfectly reflect the REAL power of the candidate in the state.

The following for example, is my evaluation of the GOP real strength in TX and FL from 1988 to 2000

Florida:

2000 - 48.8%, Adjusted: 49.1%
1996 - 42.3%, Adjusted: 52.3%
1992 - 40.9%, Adjusted: 53.6%
1988 - 60.9%, Adjusted: 57.1%

The same figures for TX are:

2000 - 59.5%, Adjusted: 59.5%
1996 - 48.8%, Adjusted: 56.7%
1992 - 40.6%, Adjusted: 54.4%
1988 - 56.0%, Adjusted: 52.2%

What are these adjustments, and why and how they are done?

From the standpoint of election specialists, sociologists and statisticians, the 2000 elections were a perfect lab, because they were ‘natural’ and accurately  reflected the diversity and the demography of America. The corrections that had to be taken were minor: adding the Buchanan votes to Bush, and adding the Nader votes to Gore.

You can assert, therefore, that the ‘natural’ power of the Reps in NY, for example, is 36% and in TX – 59.5% etc.

Things become more complicated in elections which are ‘unnatural’.

There are two reasons why elections become ‘unnatural’:

1 – A nationwide margin of one candidate over the other
2 – The participation of a third party (a serious one)

Despite these two statistical ‘noises’, you can simulate and estimate what would have been the outcomes in a specific state, had the elections been ‘natural’.

The two ‘noises’ that I mentioned, existed in the 1996 and 1992 elections.  

In 1996, the nationwide margin of Clinton over Doll was close to 9%, and Ross Perot did participate. It is agreed among the pundits, that without Perot in the race, the votes for him would have been evenly divided between Clinton and Doll.  So, if you want to ‘clear the noises’, do the following: add 4.5% to Doll (4.5 is half of the global nationwide Clinton’s victory margin). In addition, add to Doll half of the Perot’s percentage in the specific state.

In TX, for example, Perot got 6.8% and Doll– 48.8%. The adjusted (or the ‘natural’ if you will) electoral power of the Reps in TX in 1996 would therefore be:
48.8 + 9.0/2 + 6.8/2 = 56.7%.

In 1992 the methodology would be the same, but with different numbers certainly.
The Clinton’s nationwide victory margin was 5.6%. In TX Perot got 22% and Bush Sr. got 40.6%, so the Reps’ adjusted electoral power in 1992 in TX would be:
40.6 + 5.6/2 + 22/2 = 54.4%.

When we apply the same methodology to FL, we get the numbers that I brought above. If the election would be balanced (like in 1960, 1976 and 2000) then the most accurate prediction for the GOP in FL would be 46.4% since they dropped 8% in 12 years and therefore 2.7% in 4 years. Using the same methodology in TX, The best prediction for Bush in TX would be 61.8%

What we clearly see here is that in TX, the Reps are gradually becoming stronger and stronger, while in FL it is exactly in the opposite direction. What common to these two states is, that the rate of the population growth there is higher than the US’s population rate of growth?
If I were a republican strategist, I would become worried about FL.

Again: all of the above becomes irrelevant if the elections would not be balanced.

[Lunar]

I said this in your last thread but Texas is an extremely poor choice to use.  It was the homestate for Bush 41, Bush 43, and Perot.  Effectively disqualifying it for 2004, 2000, 1996, 1992, and 1988.

[Shira]

The figurs for TX are the expected results if Perot did not participate and the nationwide outcome was balanced.

[Gustaf]

You can't actually know that swings would be uniform. For instance, Perot-voters in some states might be more Clinton-friendly than those in other states.

[Shira]

Sorry, but the following was omitted in my original post.

Making the adjustments on the 1988 elections is simpler. All you have to do is to subtract 3.8% from the Bush’s percentage in the state.  3.8% is half of the nationwide margin of victory of Bush over Dukakis. Bush in TX got 56% and after adjustment it becomes 56 – 3.8 = 52.2%.

[Lunar]

The figurs for TX are the expected results if Perot did not participate and the nationwide outcome was balanced.

Yes, but you still ignore the fact that Texas is both of the Bush's homestate.  Candidates always do better in their homestate.

[Shira]

"You can't actually know that swings would be uniform. For instance, Perot-voters in some states might be more Clinton-friendly than those in other states. "

Since we lack other data, this is the best assumption:

Perot : half and half.
In addotion,  eleminate the nationwide Clinton's victory impact, by adding half of his victory margin to the RNC.

[Gustaf]

"You can't actually know that swings would be uniform. For instance, Perot-voters in some states might be more Clinton-friendly than those in other states. "

Since we lack other data, this is the best assumption:

Perot : half and half.
In addotion,  eleminate the nationwide Clinton's victory impact, by adding half of his victory margin to the RNC.

I'm not arguing that, only pointing out that it's an assumption which is flawed, since we pretty much know it's wrong. (look at any election, swings aren't uniform. Ever.). So it's not a magic device, just another perspective that might or might not help us analyze elections.

[Shira]

" So it's not a magic device, just another perspective that might or might not help us analyze elections"

Needless to say that it is far from being accurate, but is is still the best prediction.

Take the numbers of Bush Sr in 1988 and subtruct 3.81(half his nationwide margin of victory) in the compared state. When you compare it to Bush Jr ( + Buchannan) numbers in 2000 you can asses the real shift during the 12 years from 1988 to 2000.

[The Duke]

Splitting the Perot people evenly is a flawed assumption, more Perot voters would swing Republican.

Sending all Green Party votes to Gore is also wrong, since many Green's would not have voted for either of the two candidates.

Taking all Buchanan votes and sending them to Bush is a flawed assumption, since the reform party is an actual party with an actual following.  You can't know which way they'd go.

Where did the libertarian vote go?  Split evenly?  All to Bush?  All to Gore?  In states like FL and NM where the difference between Bush and Gore was below 1,000 votes, this might be important, but there is no way you can accurately predict this.

It is a flawed assumption to say that 2000 provides a perfect model because it was so close.  The country has one party that is naturally stronger than the other, this is the way of American politics through history.  The parties are not equal, there is one major party and one minor party.  To say that an election is close, so its a better election to study is wrong.

You obviously did a lot of research here, and I don't want to demean you or anything, but your model seems to be based on a lot of speculation, and not so much on hard facts.

[MODU]

"You can't actually know that swings would be uniform. For instance, Perot-voters in some states might be more Clinton-friendly than those in other states. "

Since we lack other data, this is the best assumption:

Perot : half and half.
In addotion,  eleminate the nationwide Clinton's victory impact, by adding half of his victory margin to the RNC.

Voted Perot, voted Dole.  Would have voted Perot a second time, but his second campaign did not seem as serious as his first.

[elcorazon]

While I agree with the criticisms that your analysis is simplistic, I actually think this is a valid way to look at things and to find REAL trends.  I agree, however, that home states truly muck up the analysis, so Texas is a poor state to use as an example.  I also have a problem taking seriously any in-depth analysis that misspells the name of the Republican candidate.  It's Dole, not Doll, silly.

I hope you're right about Florida, but I suspect it isn't that simple.