If Kerry has a 52.22% poll avg on 11/1, he'll have a 99.999997% win prob.

[TruthIsAll]

The results of 15 national polls of 1000 people from a
population of 150 million registered voters is equivalent to
ONE poll of 15,000 reg. voters.

If Bush steals it with those odds against him, it would be oh
sooooo obvious.

Kerry currently has a 52.22% avg in 15 polls (two-party vote
vs. Bush)

It's clear that Kerry would win the election if it were held
today, based on the average polling results (let's ignore the
Electoral College for now).

The latest 15 poll average of the two-party vote (see below)
has  Kerry at 52.22%.  If this were the result of just ONE
non-biased poll (assuming a 1000 sample size) the WIN
probability would be 92%.

Now it is very expensive to do a poll. That's why national
pollsters sample just 1,000 voters nationwide. This sample
size produces a margin of error (MoE) of 3%. This means that
if Kerry has a poll mean of 52.22%, there is a 95%
probability that he will get between 49.22% and 55.22% of the
vote.

But we have FIFTEEN polls, not just ONE. Running 15 polls,
each with a sample size of 1,000 taken from the SAME voting
population of over 150 million registered voters, is
equivalent to running a SINGLE poll of 15,000. The current
Kerry statistical MEAN for the 15 polls is 52.22%. 

But the margin of error (MoE) around the mean is lower than
3%, because our sample size is now 15,000, not 1,000. In
fact, the MoE for a 15,000 sample size is a tight +/-0.80%.
This means there is a 95% probability that Kerry will get
between 51.42% and 53.02% of the vote (a much smaller
confidence interval) were it held today. 

Ok, but we want to know what is the probability that Kerry
will get OVER 50% of the 2-party vote (just vs. Bush),
assuming a 15,000 sample size (or, equivalently, for 15
polls, each with a 1,000 sample size)?

The answer is 99.999997%. You heard it here first.

Take a look at the following 15 individual polls. Then check
the probability matrix. This shows the probability of a Kerry
win over a range of polling averages (from 50.5% to 55.0%,
including his actual current 52.22% mean), over a range of
polling sample sizes, ranging from 1,000 to 15,000.

What's the point of all this? Simple. Kerry is in a much more
commanding position than the individual polls would indicate.
In fact, he is a SURE THING if, and its a big IF, the
election were held today.

			
Latest Poll Results	Kerry%							
Date	Poll	Kerry	Bush	2-pty	MoE	
807	TIME	51	43	0.543	3.0	
804	FOX	46	43	0.517	4.0	
801	CNN	48	48	0.500	4.0	
721	LAT	48	46	0.511	3.0	
718	PEW	46	44	0.511	3.0	
												
806	IBD	49	43	0.533	3.5	
730	CBS	49	43	0.533	3.0	
805	DEMC	52	45	0.536	3.1	
802	ABC	52	45	0.536	3.0	
730	NWK	52	44	0.542	4.0	
												
729	ZOG	48	43	0.527	3.2	
806	AP	48	45	0.516	3.5	
721	NBC	45	47	0.489	3.4	
801	ARG	49	45	0.521	3.5	
722	QPAC	46	43	0.517	2.5	
												
												
N	1000											
Mean		48.60%	44.47%	52.22%	3.31%	
Median      48.00%	44.00%	52.13%	3.20%	
												
MoE	3.31%											
Std	1.690%										

If n= sample size,
MoE =1.96 / (2*sqrt(n))									
Std (standard deviation) =.5 / sqrt(n)	



WIN PROBABILITY (AT LEAST 50% OF POP. VOTE)				
(based on sample size and mean)																			
Win Probability Matrix								

Sample Sample Std	    Percent of two-party vote			
Size	MoE	Dev	50.5%	51.0%	52.0%	52.22% 53.0% 54.0% 55.0%		

600	4.00%	2.04%	59.7%	68.8%	83.6%	86.2%	92.9%	97.5%	99.3%
1000	3.10%	1.58%	62.4%	73.6%	89.7%	92.0%	97.1%	99.4%	99.9%
2000	2.19%	1.12%	67.3%	81.4%	96.3%	97.6%	99.6%	100.0%
3000	1.79%	0.91%	70.8%	86.3%	98.6%	99.2%	99.9% 100.0% 
4000	1.55%	0.79%	73.6%	89.7%	99.4%	99.8%	100.0% 
5000	1.39%	0.71%	76.0%	92.1%	99.8%	99.9% 100.0% 

6000	1.27%	0.65%	78.1%	93.9%	99.9%	100.0% 
7000	1.17%	0.60%	79.9%	95.3%	100.0	
8000	1.10%	0.56%	81.4%	96.3%	100.0	
9000	1.03%	0.53%	82.9%	97.1%	100.0	
10000	0.98%	0.50%	84.1%	97.7%	100.0 

11000	0.93%	0.48%	85.3%	98.2%	100.0% 
12000	0.89%	0.46%	86.3%	98.6%	100.0% 
13000	0.86%	0.44%	87.3%	98.9%	100.0% 
14000	0.83%	0.42%	88.2%	99.1%	100.0% 
15000	0.80%	0.41%	89.0%	99.3%	100.0% 

[wtmusic]

1. You are a seer, after all

Your earlier post said, "if the election were held today", this one doesn't. Why?

Y'know, I'm going to keep hammering this misinformation you spew out because it encourages overconfidence and therefore is destructive to the party.

Get used to it.

[snowFLAKE]

4. Yes, I've brought up the point that polls are a measuring device

of current/past events, not a Predictor of The Future.

For example, if I were to take a thermometer and make 1000 temperature readings in the greater New York area on August 9 and 10, 2004 and come up with a temperature range of 91.3 +/- 1.7 degrees and if you were to take a thermometer and do the same in the Philadelphia region on the same days and come up a temperature range of 94.7 +/- 1.4 degrees - it could be fairly conclusively stated that Philadelphia was hotter than New York on August 9 and 10.

But, could this data be used to extrapolate that Philadelphia would be hotter than New York on August 11, 2004? Maybe - but with lower certainty.

Then, could this data be used to predict that Philadelpia would be hotter than New York on November 2, 2004? Personally, I would not be confortable at all with making such a predication, no matter how top quality the sampling measurements were on August 9 and 10.

Similarly, polling is merely a snapshot of the past. Just like an icy finger of cold air could sneak down from Canada on November 2, bypass New York, and hit Philadelphia - any number of events could transpire that would determine the election results. Once again, polling could monitor these event (almost) in real time, but would have absolutely not predictive abilities.

[TruthIsAll]

5. Oh, stop already. I am saying the poll average on 11/02.

You do not read. Make believe it is Nov. 2 and Kerry has 52.22%, as he does today.

Is that too hard for you to understand?

In all my posts, I assume a what-if. That is what forecast modeling is all about. The what-if in this case consists of two assumptions:

1. It is Nov.2.
2. Kerry has a 52.22% avg vs. Bush 47.78% (assume a two-party race, for simplicity).

Wtmusic, what's your problem? Why are you so quick to put down my work? Find fault with the math, not with me. But you cannot do that, can you?

I never knock your models. Oh, yeah, I forgot, you've never done any.

[Upfront]

6. Many Here

get a lift from your work, please continue. Most do understand your method and few are slowing down because they think it is in the bag. Bush must go!

[snowFLAKE]

10. And what baffles some of us is the Events supersede Polls

It's quite banal to state that if JFK is ahead in the polls with 52.55% on November 1 he will win the election. Based on history, polls have usually been quite accurate so that's not exactly that profound of a statement.

Similarly, if politicians receive more votes during an actual election, it is highly likely based on historical trends that they will be judged to have "won" that election. Florida 2000 showed that this highly probable (statistically speaking) result can be obliterated by Larger Events.

Also, no one has yet explained to me how, if TIA's methods have any validity - why is Howard Dean not the Presidential Nominee of the Democratic Party? After all, he was handily ahead in all polls taken three months in advance of the actual "hands-on" voting.

[Barret]

16. Yup

Don't forget the polls in 2002 either.

Poll after poll after poll showed dean cleaning house. Yet it never happened. Polls can't be trusted.

[TruthIsAll]

23. Lets put this to rest once and for all.

I have presented a mathematical probability analysis. I make no assumption about October surprises or last minute shifts.

You only state the obvious, but in so doing fail to appreciate the purpose of this model: to present a sensitivity analysis of win probabilities for a given set of 15 polling results.

No more. No less.

You call the results banal and self-evident. Perhaps they are to you. But there are many here at DU who do not fully appreciate the fact that there are outliers like CNN and FOX and NBC and AP who attempt to mislead voters into thinking that Kerry received no bounce.

The POINT of this and similar posts is just this:

DON'T BELIEVE ONE OR TWO POLLS. LOOK AT ALL THE POLLS. MINIMIZE THE THE EFFECTS OF BIASED POLLS OR JUST IGNORE THE THEM.

IF I HAVE GOTTEN THAT SIMPLE POINT ACROSS WITH MY PROBABILITY MODELS, THEN I HAVE SUCCEEDED IN THIS EXERCISE.

I would like just one of you guys talk about the MATHEMATICS: The MOE, STD DEVIATION, NORMAL DISTRIBUTION, ETC.

[Tweed]

24. I'm very curious about post #20

:hi:

[TruthIsAll]

29. I suggest you look at the polling blogs and talk to

the poly sci professors yourself.

There are many who agree with my analysis: Professor Sam Wang of Princeton for one.

Go to democrats.com for the links.

mydd.com
electoral-vote.com

Oh, and one other thing. What are your bona fides?

Any degrees? Any math courses? Any programming?

Just asking.

[snowFLAKE]

35. OK, fine - you're combining the polls into one set of numbers

And that eliminates the effects of biased polls.

So if you titled Your Threads: Statistical Analysis of Polling Data

Or something Like That - then I doubt that anyone would have a beef with you.

But when you loudly trumpet terms like "Win Probabilities" - that's when a big Red Flag is raised - if you're really trying to do a statistical analysis of the outcome of the election you need to go well beyond polling data. You need to incorporate probabilities for myriad possibilities (e.g., Will the price of oil drop to $27/bbl? Will Mr. Rove develop a Brain Tumor and have a deathbed recantation of his Evil Ways? Win the Supreme (or other) Court get involved? Etc. Etc. And then statistically extrapolate these events into the Effects on Public Opinion). I submit that's a formidable task - but a Necessary One as long as you're going to promote your work as having any relevance on who will "WIN" the election.

[Tweed]

20. Have you found one poly-sci teacher to back this up yet?

:hi:

[dorktv]

2. what Bush steal an election? Never!

[tom_paine]

3. Hate to rain on your parade, but that depends on an uncorrupted voting

system.

Imperial Amerika, like other Totalitarian Nations be they Kinder/Genlter or the more typical type, doesn't have that.

[volosong]

7. Every Vote Must Be Counted

Bottom Line: Every vote must be counted. If not, we lose big time

[Barret]

14. Exactly

The repubs will be pulling out all the stops. There will be voting fraud to one degree or another in all 50 states - bank on it.

[boxster]

8. You seem to have forgotten something quite important.

Nader.

While I wish he'd be completely irrelevant in this election, reality is that he'll make a difference. This, unfortunately, isn't a two-man race, which pretty much throws your calculations out of whack.

[Zynx]

12. On election day, he won't get much more than a percent.

[Barret]

15. This statement

is like having enough confidence to claim 99.99999%.

The fact of the matter is no one knows how much nader will pull, no one knows what election fraud the repubs will try, no one knows what will happen. You might believe nader will pull 1% of the vote, but no one has any way of knowing for sure until the election. It is extremely dangerous to just assume the numbers will work out the way they do on paper 3 months before the election.

[greekspeak]

9. Thanks again for your cogent statistics

They mean a lot to me, and do a lot to build up my confidence. It is cold, hard tangible ring onto which I can grasp. I am not sure of the quantity of work you do to create these gems of posts, but I appreciate it greatly!

[madmax]

11. Great info, thanks....by Nov. 1 - I will have an ulcer. n/t

[Barret]

13. Oh christ

99.999999% huh?

Well then, guess I don't need to vote, donate, or do anything else since a win is guaranteed. And here I was thinking there was going to be hard work involved.

I've got news for you - this election is not 99.9999% guaranteed no matter what some math says. (unless it's math in a diebold computer that is)

This goes for everyone - don't sit on your ass compiling polls drinking a coke thinking we've already won. This election is going to take hard work to win.

In the time it took you to compile this "99.9999%" probability you could have been out registering voters or doing something else useful. I'm sure doing all this math with a 99.9999% result might result in happy feelings - but that's about it. Nothing useful has been done for the campaign.

[Nevernose]

17. What part of "if the election were held today"

Do you people not understand?

[Barret]

18. Even if the election were held today

You can not claim a 99.99999% probability based on polls and assumptions. (like, for example, assuming Nader will not take any of the vote)

[TruthIsAll]

21. Regarding Nader etc...

First, of course the 99.999997% seems outrageous - but it is mathematically sound; as we take more population samples, regardless of the underlying distribution, the sample mean is normally distributed. This is the Law of Large Numbers in action. It makes sense intuitively. Look at the matrix row for the typical 1000 sample poll; the probabilities are in line with what we have come to expect; from the 50.5% average poll mean up to 55% (landslide territory). The win probabilities increase as we move down (take more samples) and across (higher sample means) the matrix.

Makes sense to me. We will have to wait until November. The key benefit: we will know right away if Bush is stealing another election. He can't overcome a 15 poll Kerry 4+ point average lead unless he cheats. There is NO WAY that he can steal the Electoral College and be down 4% in the popular vote.

True, I look at the Kerry share of the vote vs. Bush only (52.22%-47.78%). I make a simple assumption regarding Nader/other/undecided voters: that Kerry and Bush will share this vote in proportion to their current split; that is, Kerry will get 52.22%, Bush the rest. This is very conservative. Many observers fully expect Kerry to get 65-75% of the vote, as Nader supporters will mostly go to Kerry, and undecided voters generally break heavily for the challenger.

So Kerry's chances are even better, believe it or not.

[Tweed]

19. So Nader polls at zero- What a joke!

This isn't a good way to go about it at all. Didn't you post numbers like this earlier today? All these numbers do is encourage people to do nothing for the campaign because Kerry is up by so much.

[TruthIsAll]

22. Read post # 21: and think...

tia

[Zynx]

26. This is based upon what's known as the two party vote.

[Zynx]

25. I just completed a state by state analysis of polls and weighted them acco

rding to their number of voters in the last election and found, if you give Kerry 70% of undecideds which Charlie Cook said is likely, Kerry wins the two party vote 53.2% to 46.8%.

Without undecideds factored in, it comes to 47.08% for Kerry to 44.18% for Bush. In any event, Kerry wins both the popular vote and the electoral college comfortably.

[TruthIsAll]

27. Zynx, thanks for the analysis. YOU do the work, unlike tweedledo.

Your approach sounds similar to mine. Where will you post it?

No pride of authorship here. Oh, well maybe a little.

We have a common goal: to cut throught the Bushit media FOG and get the Kerry bandwagon rolling. He is winning BIG and no one knows it yet.

[Zynx]

28. Since it's in spreadsheet format, it will be difficult to post.

Can you help me with that?

[TruthIsAll]

30. Just copy and paste it to DU. That's what I do. I use Excel, too.

tia

[Zynx]

31. When I do that it doesn't come out in nice neat columns in the preview.

It comes out as a jumbled mess.

[TruthIsAll]

32. Check the Message format box before posting.

That will help line up your columns, but you will still have to tweak the Excel data format to get it just right.

I know the problem. Its a pain in the butt.

[Zynx]

33. Here's my test run.

State	2000 Vote	'04 Projected 	Bush Proj	Kerry Proj
Alabama	 1,672,551	1,839,806	1,030,291	625,534
Alaska	 285,560	314,116	        175,905	        103,658
Arizona	 1,534,113	1,687,524	810,012	        776,261
Arkansas 921,781	1,013,959	486,700	        466,421
California10,965,856	12,062,442	4,824,977	6,151,845
Colorado  1,741,365	1,915,502	919,441	        823,666
Connecticut1,459,525	1,605,478	513,753	        802,739
Delaware   327,622	360,384	        151,361	        198,211
DC	   201,894	222,083	        19,988	        190,992
Florida	   5,963,110	6,559,421	2,820,551	3,279,711
Georgia	   2,596,804	2,856,484	1,485,372	1,171,159
Hawaii	   367,951	404,746	        165,946	        194,278
Idaho	   501,621	551,783	        303,481	        137,946
Illinois   4,742,123	5,216,335	2,034,371	2,816,821
Indiana	   2,199,302	2,419,232	1,112,847	653,193
Iowa	   1,315,563	1,447,119	665,675	        694,617
Kansas	   1,072,216	1,179,438	660,485	        424,598
Kentucky   1,544,187	1,698,606	883,275	        713,414
Louisiana  1,765,602	1,942,162	1,009,924	699,178
Maine	   651,817	716,999	        315,479	        344,159
Maryland   2,025,480	2,228,028	868,931	        1,180,855
Massachusetts 2,702,984	2,973,282	891,985	        1,665,038
Michigan      4,232,711	4,655,982	1,908,953	2,421,111
Minnesota     2,438,685	2,682,554	1,207,149	1,260,800
Mississippi   994,926	1,094,419	667,595	        328,326
Missouri      2,359,892	2,595,881	1,246,023	1,271,982
Montana	      410,997	452,097	        239,611	        149,192
Nebraska      697,019	766,721	        475,367	        253,018
Nevada	      608,970	669,867	        308,139	        301,440
New Hampshire 569,081	625,989	        262,915	        306,735
New Jersey    3,187,226	3,505,949	1,156,963	1,858,153
New Mexico    598,605	658,466	        283,140	        329,233
New York    6,822,451	7,504,696	2,251,409	4,352,724
North Carolina2,911,262	3,202,388	1,633,218	1,409,051
North Dakota  288,267	317,094	        193,427	        104,641
Ohio	    4,705,457	5,176,003	2,329,201	2,380,961
Oklahoma    1,234,229	1,357,652	801,015	        475,178
Oregon	    1,533,968	1,687,365	708,693	        843,682
Pennsylvania4,913,119	5,404,431	2,215,817	2,864,348
Rhode Island 409,112	450,023	        112,506	        220,511
South Carolina1,383,777	1,522,155	776,299	        669,748
South Dakota	316,269	347,896	177,427	121,764
Tennessee   2,076,181	2,283,799	1,096,224	1,050,548
Texas	    6,407,637	7,048,401	3,876,620	2,607,908
Utah	    770,754	847,829	        568,046	        186,522
Vermont	    294,308	323,739	        116,546	        165,107
Virginia    2,739,447	3,013,392	1,446,428	1,356,026
Washington  2,488,745	2,737,620	1,231,929	1,450,938
West Virginia648,124	712,936	        313,692	        342,209
Wisconsin   2,598,607	2,858,468	1,314,895	1,343,480
Wyoming	    218,351	240,186	        163,327	        67,252
Vote Totals 105,417,204	115,958,924	51,233,322	54,606,882
		Vote Percents	        44.18%	        47.09%
			Total committed 	        91.27%
			Undecided	                8.73%

Undecided	Bush Adjusted	Kerry Adjusted
183,981	        1,085,486	754,321
34,553	         186,271	127,845
101,251	         840,387	847,137
60,838	         504,952	509,007
1,085,620	 5,150,663	6,911,779
172,395	         971,159	944,342
288,986	         600,449	1,005,029
10,812	         154,605	205,779
11,104	          23,319	198,765
459,159	        2,958,299	3,601,122
199,954	        1,545,358	1,311,126
44,522	         179,303	225,444
110,357	         336,588	215,195
365,143	        2,143,914	3,072,421
653,193	        1,308,805	1,110,428
86,827	        691,723	        755,396
94,355	        688,792	        490,646
101,916	        913,850	        784,756
233,059	        1,079,842	862,320
57,360	        332,687	        384,311
178,242	        922,404	        1,305,624
416,260	        1,016,863	1,956,420
325,919 	2,006,728	2,649,254
214,604 	1,271,530	1,411,023
98,498  	697,145	        397,274
77,876	        1,269,386	1,326,495
63,294	        258,599	        193,497
38,336	        486,868	        279,853
60,288	        326,225	        343,642
56,339	        279,817	        346,172
490,833	        1,304,213	2,201,736
46,093	        296,968	        361,498
900,564	        2,521,578	4,983,118
160,119	        1,681,254	1,521,134
19,026	        199,135	        117,959
465,840	        2,468,953	2,707,049
81,459	        825,452	        532,200
134,989	        749,190	        938,175
324,266	        2,313,096	3,091,334
117,006	        147,608	        302,416
76,108	        799,131 	723,023
48,705	        192,039 	155,857
137,028	        1,137,332	1,146,467
563,872	        4,045,782	3,002,619
93,261	        596,024 	251,805
42,086	        129,172 	194,567
210,937	        1,509,709	1,503,682
54,752	        1,248,354	1,489,265
57,035	          330,802	382,134
200,093	          1,374,923	1,483,545
9,607	            166,209	73,977
10,118,720	54,268,938	61,689,986
Adjusted Totals	46.80%	53.20%

Based on the assumption that Charlie Cook used that Kerry gets
70% of undecideds. 

Turnout was based on 10% higher turnout than 2000, hence about
10,000,000 more voters.

[Zynx]

34. Thanks btw.

[jmknapp]

36. More realistically...

Check this web page:
ELECTORAL ODDS CALCULATOR

Using the default assumptions, Kerry has an 81% chance.

Change Kerry's odds of winning various battleground states and recalculate, but there is no way this election is a foregone conclusion.

If FL and OH go Republican (a DISTINCT possibility), Kerry probably loses.