10/9 ELECTION MODEL: CHARTING THE RACE

[TruthIsAll]

http://www.geocities.com/electionmodel /


Today the Zogby daily tracking poll has Kerry leading by 46%-45%.

Kerry averaged 326 EV in the Monte Carlo simulation of 5000 election trials based on the latest state polls. He won 4938 of them, so his win probability is past body temperature: 98.76%.


He’s projected to win 51.27% of the popular (state-weighted) vote


The base case assumption is that Kerry gets 60% of the undecided/other vote. If he gets 75%, his expected EV jumps to 340 with a 99.9% win probability. If he splits the undecided with Bush (50%), his expected EV drops to 315 with a 96.0% win probability.



Kerry leads in 15 of 21 Battleground states shown in this chart:



This chart displays Kerry's current and projected probabilities of winning the Battleground states:



Kerry’s average is rising for the group of 9 national Independent national pollsters: Zogby, Harris, ARG, IBD, The Economist, ICG, IBD/CSM, Quinnipiac, Pew. In this group, Kerry has an 84.4% probability of winning over 50% of the popular vote:



In the Corporate and Independent National group (18 polls), Kerry's win probability has risen to 93.6%
The National and State EV models are in sync once again.




Bush job approval is UNDER 50%. No candidate has ever won the presidency with an average approval rating below 50%:



This chart shows that Kerry's win probability based on an average of several polls is higher than his win probability calculated for a single poll for the same numbers. The margin of error is reduced as when more people are polled.

The MoE for a single national poll of 1000 individuals is 3.10%.
The MoE for a 9 national poll group of 9000 is 1.03%.
The MoE for an 18 national poll group of 18000 is 0.73%




Monte Carlo simulation is used to calculate the probability of a Kerry win based on current state polls. If Kerry wins X out of 5000 simulated trial elections, his probability of a victory is: P = X / 5000. Kerry had an average of 326 electoral votes in today’s 5000 simulated election trials. His win probability is P = 98.8%.

This chart shows the distribution of simulation trial outcomes:



As you can see from this chart, the EV simulation outcomes cluster in the 320-340 range:





http://www.geocities.com/electionmodel /

[Fluffdaddy]

1. I don't think there are any undecided votes left. Nader is the Killer

Lets hope the 1-3% of people that "May" vote for him will wake the F-up an not waste their vote

[TruthIsAll]

3. Of course there are undecideds, otherwise the poll would total 100%

Based on the poll averages, there are 6-7% undecided/Nader voters.
And Nader will get less than 1% this time.

[TruthIsAll]

8. Insert Zogby's numbers into Electoral-vote.com today and ..

you will see that Kerry has over 320 EV.

Survey-USA?
Gallup?
Strategic Vision?
Mason-Dixon?


Trust Zogby. Trust ARG.

Don't be bamboozled by Repub pollsters.

http://www.geocities.com/electionmodel/

[farmbo]

2. Great post...thanks for the low-down...

...I wish the election was THIS Tuesday.

:headbang:

[TruthIsAll]

4. We don't want the election today, let Kerry build a lead so large

that Bush can't steal it.

[TruthIsAll]

5. The Bush debate meltdown will bring his numbers down further.

Last night's debate will have profound repercussions.

I fully expect Kerry will move to 340 EV and 52% of the vote.

And that will give him a 99.9% win probability.

tia

[KharmaTrain]

7. Great Work

Too bad there's really no efficient way to work "unregistered" voters into this mix. While Democrats have won this battle in recent months, RoveCo. has been at it for the past four years, so I'm not certain the ratio here.

[Selwynn]

6. This stuff rocks -- IF bush wins though, I am going to ask you...

..what went wrong.

Not in an accusing way, but I am fascinating by the variables. If Bush wins, I'll want to figure out what it was in reality that did not behave like the model predicts, and learn from that.

Right now, it looks like great news! Keep it comin.

You know what would be an interesting experiment - find data on the polls from the 2000 race, say october 2000, and to the same kind of statistical analysis and see what it predicts.

[TruthIsAll]

9. I am not PREDICTING, just projecting based on undecided allocations.

I use the latest state and national polls, average them, project the undecided votes to Kerry and calculate probabilities of winning the states and the national vote.

This is NOT an econometric model.

The purpose of the model is to give an accurate picture of where we have been and where we are, based on the myriad of good and bad polls out there.

My only assumption is that Kerry will win the undecided.

No multiple regression, no exponential smoothing. Nada.

Just polling stats, random numbers, MoE's.

And the normal distribution to compute probabilities of winning each state and the national vote based on these polls - and my undecided assumption.