2004 User Predictions - Discussion

[CARLHAYDEN]

This topic is to re-start the discussions around the user predictions located at the 2004 Prediction page.  I have created another topic to discuss the technical issues with the feature.

The 2004 presidenial election is one of the most difficult to predict for a variety of reasons.

First, the normal 36 year cycle is NOT occuring largely due to demographic changes over the last forty years.  Among those changes are:
 
Longevity has increased even as the birth rate has dropped.  The result is that the actual electorate (those who do vote), as opposed to the voting age population (VAP), is unusually high by historical standards.  

Further, due to illegal immigration, a significant proportion of the VAP is ineligible (legally) to vote.

Additionally, the percentage of the VAP which is disqualified for voting based on felony convictions (where full civil rights have not been restored) is at a historical high.

Second, the public opinion polls available to the general public are more defective today than at any time in over fifty years for a couple of reasons:

Some 'poll' are intentionally designed to return the result desired by the entity paying for the 'poll'.  Such polls include, LA Times, CBS, Princeton Associations and some state polls (in Arizona, my home state, the 'crank' poll is notorious for overstating Democrat preference).

Other polls are unintentionally slanted either in reaction to the suprisijng 2000 results (a number of polls significantly increased the black percentage of respondents in their surveys in reaction to that election) or because of the changes in the telecommunications industry (an analogy would be if you only polled people who watch over the air television only to determine television viewing preferences, and excluded those who watch cable or satellite television).

In my view, the key data to examine:

(a)  the economic statistics (which is a good, abeit imperfect indicator),

(b) partisan voter registration data (available in most states outside the south and border area), which has been a very good indicator of election results,

(c) the 2002 and 2003 election results (a fair, abeit imperfect indicator),

(d) nomination data (a good indicator),

(e) technical campaign data (fundraising, organizational efforts, campaign coherency, etc.),
which are good indicators in close elections, but sometimes difficult to quantify.

(f) likeability (a good indicator in close elections). and

(g) historical voting patterns (a very good indicator except in lanslide elections).

As has previously been noted, if the economy continues along the path established in the last two quarters, every president in the last fifty years with such statistics seeking reelection has been reelected.

If you check the voter registration statistics (available in the links to the states), you will find that in more than ninety per cent of the states with partisan registration (this includes most of the states), the Republicans have improved their relative ratio to the Democrats.

In the 2002 and 2003 elections, the Republicans did far better than expected.  They not only retained control of the House, they regained control of the Senate, and did far better than expected in the Gubenatorial elections.

Every President who has sought reelection without serious challenge in his own party over the past half century has been reelected (Clinton in 1996, Reagan in 1984, Nixon in 1972, Johnson in 1964 and Eisenhower in 1956).  Every president who sought reelection and faced serious challenge within his party lost (Bush in 1992, Carter in 1980, and Ford in 1976).

While Kerry did far better in fundraising than is usually the case for challengers, Bush raised more money than Kerry.  Moreover, Kerry has not really gotten his act together organiztionally while Bush has one of the best organized campaigns in recent years.  Further, while Bush has a relatively coherent campaign, Kerry has a major difficulty trying to square his record in the Senate with positions palatable to enough voters to be elected.

In the end, where major trends are not going in one particular way and the candidates are not seen as unqualified, likeability has a significant factor in ultimate voting decision.  Every source I have seen shows that the voting public finds Bush more likeable than Kerry.

Finally, except in landslide elections (and I have yet to see any credible source suggest that 2004 is likely to be a landslide), historical voting results give a good project future results (given sophisticated modeling).  Please note that the last Democrat nominee to win more than half the vote was Carter in 1976 (with just 50.08% of the vote).  Since that election Republican nominees have exceeded fifty per cent of the vote on three occasions (1980, 1984, and 1988).  So, those who suggest that Kerry will do better than Carter did in 1976 are, IMHO, blowing smoke.

Given these points of analysis, I project that, the total presidential popular vote will total approximately 108,200.000, and that Bush will receive 51.10 % of that vote, Kerry will receive 44.97 % of that vote, Nader will receive 1.51 % of the vote, Badnarik 1.06 % of the vote, Cobb will receive 0.71 % of the vote, Peroutka will receive 0.43 % of the vote, with other candidates (the usual menagarie) approximatly 0.22 % of the vote.

My electoral college map is posted in the predictions section.  A number of states are either marginal or lean to one candidate.  However, the math is such that it is virtually impossible (unless a major catastrophe occurs in the next three and half months) for either the electoral college or the popular vote to go for Kerry.

I fully expect the partisans of Kerry, and those on the left in general to howl about this posting, but I suggest they get it out of their system now, as it will reduce their frustration on the night of the election when the cold hard data is accumulated.

[CARLHAYDEN]

Notice, that except for my prediction, I cited facts.

If we want to go into "would have." I suggest that if Sen. Gramm's campaign had not screwed up and alienated the Govenor of Louisiana, he would not have lost the contest there to Buchanan.  He would have been a far more formidable candidate than Dole.  

Dole was a lousy candidate (Republicans, when not in front of a reporter, generally referred to him contemptously as "senator taxman").  He also alienated social conservatives.  

The fact is that the people who voted for Perot could have voted for Clinton, and chose not to so cast their vote.  

Speculation is interesting, but facts are more reliable.

[CARLHAYDEN]

Its really hard to say.

In 1992, a lot of conservatives were very angry with Bush.

They didn't want to vote for Clinton, but they wanted to punish Bush for his betrayal.

Suggests that many of the Perot voters would have declined to vote in the Presidential election (a few would have voted for the Libertarian candidate).

Note, that while Bush was getting a smaller percentage of the popular vote than Goldwater (or any other Republican candidate in fifty years), Republicans had a net gain in House seats.

[CARLHAYDEN]

First, thanks for the comments.

Second, I do agree that Clinton would probably have gotten slightly over fifty per cent without Perot in 1996, but he would not have gotten a majority of the Perot vote, and his reelection percentage would have been lower than Reagan in 84, Nixon in 72, Johnson in 64, and, Eisenhower in 56.

However, it would have been the lowest sucessful reelection percentage since 1948 (when Truman also didn't get a majority of the vote).

[CARLHAYDEN]

In general, quite plausible. however, I believe New Mexico is more likely to vote Republican than Washington, and Iowa more likely than Michigan.

The bottom line is that Bush has the hearland advantage.

[CARLHAYDEN]

In general, quite plausible. however, I believe New Mexico is more likely to vote Republican than Washington, and Iowa more likely than Michigan.

The bottom line is that Bush has the hearland advantage.

I need the order in which the states are called on election night to make a proper analysis. If you can provide that I would be grateful.

The "order" in which states are "called" by the television networks will depend upon a number of factors.

First, will they wait until the polls are closed in a state before calling it, as they have pledged?  As you know, Florida was "called" for Gore before polls had closed in the panhandle by a couple of networks.

Second, it depends upon how good the technical analysis is of the data used by a particular network.  

Third, while the networks got 'egg' on their faces for their error in Florida four years ago, there always is a tendency to want to be 'first.'

Fourth, don't be concerned by the first hour to two hours of raw vote as the initial vote will be coming in from the eastern time zone (Kerry's strongest) and from the big cities (rural votes come in latter.

Another factor to consider in the raw vote totals is whether it includes 'absentee' voting, which is looming larger in every voting cycle.  Some jurisdictions count those votes first (since they're already avaialable at the central facilities), while others count them last (alleging delays in verification), and others in the middle (when the machines have excess counting capacity.

Expect the following jurisdictions to be "called" for Kerry about a minute after the polls close:

District of Columbia
Hawaii
Massachusetts
Rhode Island

Expect the following jurisdictions to be "called" for Bush about two minutes after their polls close:

Alaska
Idaho
Kansas
Nebraska
North Dakota
Oklahoma
Utah
Wyoming

Note the the bulk of Kerry "quick" victories are in the Eastern time zone, whereas Bush's "quick" vitories are in the Central/Western and latter (Alaska) time zones.

Its one of the few times I watch a few minutes of Dan Rather is election night.

Its funny to watch Rather bubbling over joyously announcing the early Democrat victories, and then becoming more and more glum as tallies from the hearland roll in to bury his delusions.

[CARLHAYDEN]

I believe it starts

Indiana
Kentucky
Florida

You asked about, "calling," which means stating making a determination which way a state will actually end up supporting, not when the election polls have closed.

BTW, since Florida is in two time zones, the closing time for polling places differs (after their last fiasco, the networks are going to treat Florida with kid gloves).

I expect that both Indiana and Kentucy will be 'called' by the networks for Bush about a five minutes after the polls close in those states.

[CARLHAYDEN]

Oh, that's simple.

The closing time on polls is earlier in Indiana and Kentucky than Rhode Island.

This 'closing time' is historical and dates back to around 1910 when both Indiana and Kentucky were largely rural.

Generally, farmers were 'early to bed and early to rise,' so the polling times followed their schedule.

Rhode Island by contrast was largely urban and industrial at the time, so later polling hours (closing times) were provided so that industrial laborers could cast their ballots after they left work.

As an interesting side light, in New Hampshire, a township by the name of Dixville Notch, accounces its vote at approximately 1 a.m. on Tuesday morning.

[CARLHAYDEN]

1992:



Clinton's lead INCREASES by 2% after Perot drops out the first time.



Clinton's lead remains at 16% after Perot re-enters.

Also, as Perot climbs, Clinton's lead decreases, and as Perot tapers off at the end, Clinton's lead climbs.

I just don't see any evidence Perot hurt Bush in the polls.  The only argument you can make is if that you say Perot's ads hurt Bush, but there really is no evidence to support that claim.

I think we can safely say Perot had little effect.

Well Boss, thanks for pointing out an example of the inaccuracy of a major poll.

While Gallup had Clinton at 49 in their closing poll, he actually received 43 per cent of the actual vote (a 6 point difference).

Gallup had Perot at 14 at closing whereas he received nineteen per cent of the popular vote, a five point differenct.

Further, when a number of studies looked at the Perot 92 vote, they found that in 1988 those same voters had split approximately as follows (minor differences in studies);

Bush 1988          -          60% of 1992 Perot votes
None in 1988        -        30% of 1992 Perot votes
Dukakis/other in 1988 - 10% of 1992 Perot voters

If you compare the combined Bush 92 and Perot 92 vote to the Bush 88 vote (go state by state, county by county if you wish) you will see a remarkable similiarity (Arkansas excepted).

The thing is that the 1992 Perot voters were very angry at Bush such that (according to the best projections I have seen), if Perot had not been on the ballot in 1992, they would have 'voted' as follows for President.

Other   10%
Clinton 20%
Bush    40%
None    30%

Further evidence of the normally pro-Republican nature of the Perot vote can be seen in the fact that although Bush in 92 had the lowest popular vote percentage of any Republican candidate since Alf Landon, Republicans actually gained in the U.S. House of Representatives.

[CARLHAYDEN]

Carl,

First, I'd like to knwo where you got your numbers from, then I will point some things out.

As I said, they come from a number of studies done by a couple of Political Science professors assisted by graduate students.

[CARLHAYDEN]

Boss,

Your projection of the 92 Democrat vote closely mirrors the actual Democrat vote in both 96 and 2000,

So, it looks pretty credible to me.

[CARLHAYDEN]

Unfortunately, I do not have such number.

Interesting that you should ask, as I asked one of the persons who produced one of the studies I cited, and he stated that there just wasn't enought interest in putting the time into replicating the 1992 Perot studies.

He had done some preliminary work and guardedly suggested that had Perot not been on the 1996 ballot, his votes would have gone as follows:

Others                         5%
Clinton                       10%
No vote for President 20%
Dole                            65%

While the Perot voters did not like Dole, they weren't as angry with him as they were at Bush.

Also, the Republicans had championed a number of Perots positions.

Dole had resigned from the Senate, of which they approved.

Clinton was perceived as big government (health care, gun control, opposition to term limits, etc.)

For most of the 1996 Perot voters, they would have "held their nose" and voted for Dole.

[CARLHAYDEN]

Your maps looks about right.

The data I earlier provided was national data, and there were minor regional variations.  The data is, as I previously noted, better for 92 than 96.

Generally speaking, Perot voters in both elections were slightly more likely to back the Republican candidate in the South than elsewhere in the nation, so Florida would probably have been incredibly close in 1996.

Conversely, Perot voters in both elections were slightly more likely to back third party candidates in the West, so Nevada would have probably been incredibly close in 1996.

Perot voters in the midwest (particularly upper midwest) and northeast (particularly new england) were just a tiny bit more likely to back the Democrat candidate in both elections than nationally (probably woudn't have had a significant impact on any state).\

[CARLHAYDEN]

Actually, it depends upon you estimate of turnout.

If you expect heavy turnout in the black belt counties and normal turnout in the ozarks, then you could show Arkansas as competitive.

A normal turnout turnout would give Bush a clear lead.

[CARLHAYDEN]

In the post to which I previously refered you to, I briefly listed numerous criteria which I used in developing my projection.

Due to limitations on the length of posts, I did not elaborate on the methodology of each of these criteria.

I will begin listed them in order, starting with the key economic statistic, "unemployment rate."

If you review the unemployment rate statistics back to ther 1932 election (they are not available for prior Presidential elections), you will see that every single President renominated by his party forreelection with an unemployment rate of less that 7.1% has been reelected.

In those Presidential elections during the period involved where the incumbent was renominated and had an unemployment rate of 7.1% or higher, in a majority of the cases, they were defeated (the exceptions are FDR in 36 and 40, and Reagan in 84).

Based on a statistical analysis of this data, it seems to me that Bush would be reelected by a margin of 4.48% or higher.

[CARLHAYDEN]

In the post to which I previously refered you to, I briefly listed numerous criteria which I used in developing my projection.

Due to limitations on the length of posts, I did not elaborate on the methodology of each of these criteria.

I will begin listed them in order, starting with the key economic statistic, "unemployment rate."

If you review the unemployment rate statistics back to ther 1932 election (they are not available for prior Presidential elections), you will see that every single President renominated by his party forreelection with an unemployment rate of less that 7.1% has been reelected.

In those Presidential elections during the period involved where the incumbent was renominated and had an unemployment rate of 7.1% or higher, in a majority of the cases, they were defeated (the exceptions are FDR in 36 and 40, and Reagan in 84).

Based on a statistical analysis of this data, it seems to me that Bush would be reelected by a margin of 4.48% or higher.

How does this change if you use change in unemployment rate instead of the unemployment rate?

Good question.

Even though the unemployment rates in 36, 40 and 84 were of a magnitude to disturb the electorate, the unemployment rates were NOT increasing in the year before the election, and were generally decreasing.  In short, the incumbents in those years (FDR and Reagan) were NOT blamed for the relatively high unemployment rates.

[CARLHAYDEN]

Wouldn't this figure have to be lowered due to Welfare to Work (ie, lower official unemployment figures when the problem is of the same magnitude)? In other words, are unemployment figures pre-Welfare to Work strictly comparable to those afterwards...
I'll have the same complaint when you mention post-91 GDP data, so be prepared.

In the real world, no data is perfect.

Please note that I did NOT base my prediction solely on this particular factor.

You approach seems to be that unless data is perfect its worthless.

I suggest that the data speaks for itself for those who are reasonable.

[CARLHAYDEN]

Wouldn't this figure have to be lowered due to Welfare to Work (ie, lower official unemployment figures when the problem is of the same magnitude)? In other words, are unemployment figures pre-Welfare to Work strictly comparable to those afterwards...
I'll have the same complaint when you mention post-91 GDP data, so be prepared.

In the real world, no data is perfect.

Please note that I did NOT base my prediction solely on this particular factor.

You approach seems to be that unless data is perfect its worthless.

I suggest that the data speaks for itself for those who are reasonable.

Yeah, I know that...imperfect data isn't worthless o/c, but mathematical models based on imperfect data are likely to produce misleading results - and have the potential to produce entirely bogus results. Of course, such minor problems may cancel each other out - they'd have to be tugging in all directions though.

Virtually all data used in the social sciences is imperfect, as such absolute precision is impossible.

However, one can make projections with the data with a relatively high confidence level when other independent variables closely correspond with the projections made by the methodology.

BTW, instead of double talk, please cite me one example of an incumbent President seeking reelection when the unemployment rate (as measured by the Labor Department) was below 6%.

[CARLHAYDEN]

Still waiting for the example of a President seeking reelection who was defeated when the unemployment rate was less than six per cent.

[CARLHAYDEN]

Still waiting for the example of a President seeking reelection who was defeated when the unemployment rate was less than six per cent.

Still waiting for the example of an incumbent president who won when the Redskins lost their final home game before the election.

I don't recall EVER alledging a football based prediction.

I DID use as part of my prediction the unemployment rate.

The poster to whom I had responded tried to argue imperfection of data.

I had given him numerous examples of the proof of my method, and he provided no contrary examples.

So, since you entered the discussion, do you have ANY examples of an imcumbent president seeking reelection who was defeated when the unemployment rate (as measured by the Labor department) was 6% or lower?

[CARLHAYDEN]

You seem to be losing it.

Your post is totally non-responsive.

I posted a question to you to which you have yet to respond.

BossTweed posted a response to my post to you which had nothing to do with the previous question.

I realize one of the techniques of the left is to try to change the subject.

Try answering the question in the first place instead of trying to change the subject.

[CARLHAYDEN]

Virtually all data used in the social sciences is imperfect, as such absolute precision is impossible.

However, one can make projections with the data with a relatively high confidence level when other independent variables closely correspond with the projections made by the methodology.

BTW, instead of double talk, please cite me one example of an incumbent President seeking reelection when the unemployment rate (as measured by the Labor Department) was below 6%.

Yeah, sure.
a) Yes, exactly. That's why mathematical models using them should be treated as nice ideas that may even point something worthwhile out, but not as portents of the truth.
Anyways, your prediction is not based solely on that, nobody has claimed that.
b) In plain English? Which independent variables?
c) "Double Talk"? Me? Never.
As you word the question, 1996. However, I think you meant one seeking it and losing. I guess you'd have to go back to 1912 or 1892. This is *very much* a change of topic, btw. I know you're a registered Democrat but claiming yourself as a member of the left is more than just a bit thick.

Yes, my post was unresponsive. That's because yours was.

What miami was saying is, of course, "you can always find some chance unrelated event that looks like it's related". I don't agree that unemployment rates have no effect on elections (of course not...I'm from Germany), but "I don't recall ever alleging a Football based prediction" is a very weird way of answering that if answering it was indeed what you were trying to do. Don't you agree?

BTW: Part of this whole irritation is this: As your prediction is obviously off (why? because I say so ) my reply was pointing out possible causes of the error, something I never actually said. That much is my fault...

First, don't know if you're doing drugs or simply flipping out.

Second, I posted a methodology (specifically stated) where Zero exceptions (the unemployment model).  You derided it by alledging the unemployment data was insufficently accurate.  You provided NO example of where the model was wrong!

Third, mathematical models have proven their worth over the years and are more than just "nice ideas."

Fourth, if you went back to my original post, you would see that I used a number of factors (explicitly stated) to arrive at my result.  To cite just one of them, no incumbent President seeking reelection has been defeated unless he had significant primary opposition for the period going back to the end of the second world war (Bush had no significant primary opposition).

Fifth, with respect to your "word question," I can only wonder if you are referring to a post by another poster (perhaps BossTweed).  I never had a 'word question' on this thread.

Sixth, where do you get the idea I ever claimed "myself of the member of the left"?!?

Seventh, my notation that I had never used a football prediction was a response to BossTweed.

Eighth, so your ultimate basis for disagreeing with my prediction is simply "because I say so"?

Suggestion, go back and reread the posts on this thread.