THE TRUTH ! 16 NATIONAL POLLS - KERRY PROBABILITY OF WINNING IS 91%-99%

[TruthIsAll]

There has been misinformation put out by a number of sources
(AP, MSNBC, FOX, CNN) regarding Kerry's current poll numbers
and his prospects of winning. 

Hopefully this will clear things up.

KERRY IS AHEAD IN VIRTUALLY EVERY NATIONAL POLL. HIS
PROBABILITY OF WINNING, BASED ON PROJECTING HIS VOTING % FOR
EACH POLL AND AVERAGING THE RESULTS RANGES FROM 91-99%.

THE MOST LIKELY SCENARIO GIVES HIM A 96.89% PROBABILITY OF
WINNING, ASSUMING HE GETS 70% OF THE OTHER/UNDECIDED VOTE.

My 7/24 Kerry EV simulation analysis shows he has a 98.4%
probability of winning the election, based on probabilities of
winning each state.
A National Polling Analysis was also included.

This analysis is also based on more complete latest National
polling numbers. I have included 16 polls from
www.pollingreport.com

The analysis is fairly straightforward.

1. I project Kerry's final result by adding an allocation
adjustment for undecided/other votes, for the range of
60-100%. This way, we can see the sensitivity of the
projections and associated probabilities to the allocation.

2. For each projection (allocation) set I calculate the MEAN
and STANDARD DEVIATION (STD) of the projections.

3. The  STD, MEAN and .50 winning threshold are input to the
Cumulative Normal Distribution function, which returns the
probability that Kerry will get a MINIMUM of 50% of the
POPULAR vote and win the election, assuming he gets a minimum 
of 270 Electoral votes.

Important note:
The STD for Bush is 1.83, as opposed to 2.81 for Kerry. This
indicates that there is less VARIABLITY in the Bush poll
numbers then for Kerry, as they are clustered more closely
around the 44.88 Bush average (MEAN) for the 16 polls.

This could also be an indication that Bush is pretty much, at
least for now, STUCK AT 45%, with LIMITED upward potential.




			Projected Kerry % of Vote				
Latest Polls	Allocation of other/undecided to Kerry					
Poll	Kerry	Bush	60%	70%	80%	90%	100%
time	50	45	53.00	53.50	54.00	54.50	55.00
fox	45	44	51.60	52.70	53.80	54.90	56.00
cnn/gal	49	47	51.40	51.80	52.20	52.60	53.00
lat	48	46	51.60	52.20	52.80	53.40	54.00
pew	46	44	52.00	53.00	54.00	55.00	56.00
							
ibd	44	41	53.00	54.50	56.00	57.50	59.00
cbs	49	44	53.20	53.90	54.60	55.30	56.00
demc	52	45	53.80	54.10	54.40	54.70	55.00
wp	46	46	50.80	51.60	52.40	53.20	54.00
nwk	51	45	53.40	53.80	54.20	54.60	55.00
							
zogby	46	44	52.00	53.00	54.00	55.00	56.00
ap	45	49	48.60	49.20	49.80	50.40	51.00
nbc1	54	43	55.80	56.10	56.40	56.70	57.00
nbc2	45	47	49.80	50.60	51.40	52.20	53.00
arg	49	45	52.60	53.20	53.80	54.40	55.00
qnpiac	46	43	52.60	53.70	54.80	55.90	57.00

Average	47.81	44.88	52.20	52.93	53.66	54.39	55.13
Stddev	2.81	1.83	1.61	1.57	1.60	1.69	1.83
Prob Kerry win 		91.36%	96.89%	98.90%	99.54%	99.74%
							
Notes:							
nbc1 - Princeton Associates							
nbc2 - Wall Street Journal							

[Kazak]

1. It's simple...

polls or no polls, we must turn out in droves.

[dbt]

2. Nominate this post for the home page!

I just did. TIA been doing us all a great service with these projections, backed up by some expertise in statistics.

:bounce:
dbt

[snowFLAKE]

3. The only problem is that a statistical approach doesn't apply

To predict the outcome of The Election.

Simply because it is subject to forces that have not, indeed cannot, be incorporated into the polling data.

I will attempt a simple-minded analogy to explain. Let's say TIA has observed the route I take when driving home from work for the past four years. Based on these observations, he has seen that I take "Route A" 997 times and "Route B" 3 times. He or she therefore predicts that there is a 99.7% chance that I will take Route A on November 2, 2004. However, it turns out that on November 2, 2004 my spouse is returning from a journey and I must make a detour to the airport on my way home, thus taking "Route C."

I am sure that Rove & Co. will do everything they can - for example in the form of the much anticipated "October Surprise" - to make sure that "Route A" is not followed on election day. Whether they will be successful or not is hard to predict statistically - for instance, such a strategy didnt' work for Jimmy Carter and his attempt to rescue the Iran hostages. However, I suspect the chance of success is higher than 1-11% as projected by TIA.

[TruthIsAll]

5. Tell me something I don't know. You're mantra is almost cliche.

Of course, we don't know what Rove might try to do.

But know this: Unless the election is outright stolen AGAIN via computer hacking or another coup d'etat or terrorist attack, Kerry will win.

I use the most current polls information available, which you won't get from the talking heads who lie about Bush being in the lead. I just want to clear the air and present the FACTS and a reasonable STATISTICAL analysis of those FACTS.

You will find other election forecasts biased and off the mark on the Net as well as from the cablewhores.

I am using Probability and Statistics 101 methodology. How is it our so-called experts in the media won't use the latest data or apply the straightforward, proven statistical methodology?

[snowFLAKE]

8. OK, but please explain why Mr. Dean isn't the Democratic Nominee

Because, by using your "predictive" methods, virtually all polls done over a period of several months in advance of the Primaries had him as The Sure Thing.

The point is that polls have no predictive abilities statistically speaking. Sure, you can attempt to extrapolate into the future based on trends from past elections you feel might mimic this one, but then you've moved from being completely objective (i.e., statistical) to being subjective.

Presidential elections are unique events not amenable to statistical methods. Unless, of course, you account for all eventualities - will Mr. Kerry self-destruct ala Senator's Hart's Donna Rice affair? What's the probably of that? Will Mr. Bush finally snap in public? Will the raping-the-children-in-Iraq story finally get some media attention? Will anyone care? Again, what's the probability of these events? Basically, they haven't happened enough times to assign numbers.

[Paradise]

26. Okay, however I just want to know what the polls say,

truthfully, at the points in time they are taken.
It's the lies about those polls, the slants and the manipulations, that make me crazy. :crazy:

[SheilaT]

4. Two comments:

One is that no sitting president has ever been re-elected with poll numbers this bad at this point in the election cycle.

The other is that Snowflake is right, that the evil men in charge will do anything and everything possible to see to it that they stay in power. We really won't be able to rest easy until noon of January 20, 2005 and John Kerry takes the oath of office.

[TruthIsAll]

6. MSNBC Live vote seems to match the projections

Live Vote

http://www.msnbc.msn.com/id/5490309/
If the presidential election were held this week, who would you vote for? * 35246 responses


George W. Bush
45%

John Kerry
52%

Ralph Nader
2%

Not a scientifically valid survey. Click to learn more.

[Paradise]

7. Thanks TIA! They make me crazy! :( nt

[opihimoimoi]

9. The Visine Formula was strengthened secretly. Coffee was also

increased in Awakening Power.....

America is waking up...too many Clarkes, Frankens, Moores, etc.

Bushies helping by persistently fucking up in every direction they take. They missed the KISS thing, made things so complicated, their Base got lost.

[Bucky]

10. A few flaws in your methodology: 2-candidate model, MOE, popular vote...

Sorry, no sale. I agree that it looks good right now but your model only "predicts" what the outcome would be if the vote were held today. The probability is not the probability for november, but rather the confidence rate that, if you took the same polls again, you'd get results within the margin(s) of error of all these polls. I like your outcomes, but you're math isn't answering the questions you claim it is.

Besides which, you're only selecting the two candidate model. There is a third candidate who's running 4-8% in most polls right now. That will probably drop off, but we can't assume that it will.

Three months out, obviously all the players haven't played all their cards. Whatever they do in October will certainly impact the vote.

Finally, you're dealing with the popular vote as if it matters. It doesn't. Kerry could get 60% of the vote and it woldn't matter if the votes aren't arranged in the right states to produce an electoral college majority.

[TruthIsAll]

12. I don't have anything to sell.

You:

Sorry, no sale. I agree that it looks good right now but your model only "predicts" what the outcome would be if the vote were held today. The probability is not the probability for november, but rather the confidence rate that, if you took the same polls again, you'd get results within the margin(s) of error of all these polls. I like your outcomes, but you're math isn't answering the questions you claim it is.

Me:
Of course, it calculates probabilities if the election were held today according to the split in the undecided vote. That's called sensitivity analysis. Its a projection model which, like all models, will be updated with the latest data as it becomes available. Do you have a better approach?


You:
Besides which, you're only selecting the two candidate model. There is a third candidate who's running 4-8% in most polls right now. That will probably drop off, but we can't assume that it will.

Me:
I take into account the "other" voters, which includes undecided and Nader. And based on historical data, they will go to the challenger by 60-80%. Nader got 2.7% of the total vote last time; he will get less than that this time. I allocate Kerry's share of the "other" votes, from 60-100%. Again, the sensitivity analysis gives us a range of outcomes and probability of Kerry winning, depending on his share of the "other" vote.

You:
Three months out, obviously all the players haven't played all their cards. Whatever they do in October will certainly impact the vote.

Me:
No kidding. That's obvious. And that's why I will run this and other simulation models as I continue to get new data.

You:
Finally, you're dealing with the popular vote as if it matters. It doesn't. Kerry could get 60% of the vote and it woldn't matter if the votes aren't arranged in the right states to produce an electoral college majority.

Me:
Sorry to say this, but you have just revealed your total ignorance by that statement. There has never been an election where a candidate got over 51% of the vote and lost the election. That can only happen if the margin is razor thin.

When a candidate gets 55% of the popular vote, he has a 99.999% lock on more than 270 EV's probability of winning, I suggest you check out my electoral vote simulation model.

[Zidane]

15. I think you've just lost credibility

"Sorry to say this, but you have just revealed your total ignorance by that statement. There has never been an election where a candidate got over 51% of the vote and lost the election. That can only happen if the margin is razor thin."

Check out election 2000. Gore won the popular, but lost in the electoral college. If we decided president by popular vote Gore would be in office today.

And as far as the 99.999% lock... I can tell you've never been around official party strategists. Kerrys people don't take more than one look at national polls. They get polls from state to state, figure out what the percentage is, and look at the electoral vote count. There are MANY possible scenarios in which you can win the popular but lose the electoral.

[TruthIsAll]

17. I suggest you look at this.

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=132x585394

[TruthIsAll]

18. Gore won Florida. It was stolen. Remember?

How quickly you forget.

[Zidane]

22. Actually

I was talking about NATION WIDE numbers. Gore had a high popular vote nation wide - even if he had lost Florida legitimately by, say, 10 votes, he STILL would have had a higher popular vote but lost in the electoral college. Point being - it can happen.

[Paradise]

27. Yes, TIA, and the exit polls were correct, as they have always been! nt

[Zidane]

13. Exactly

"Finally, you're dealing with the popular vote as if it matters. It doesn't"

And a lot of people look at national polls as if they mean something. When they are close as they have been national polls are mostly worthless.

Instead of posting national polls people need to post STATE polls of all 50 states. i.e. New York is polling 70% kerry and 30% bush. Then you need to add the electoral votes of each state up, and figure out who has the most.

Doing that is a hell of a lot more reliable than looking at national polls.

[Zidane]

11. This doesn't seem very accurate

"THE MOST LIKELY SCENARIO GIVES HIM A 96.89% PROBABILITY OF
WINNING, ASSUMING HE GETS 70% OF THE OTHER/UNDECIDED VOTE."

*ASSUMING* he gets. Hell, I have an upper 90% shot of winning the election in november as well *assuming* I were to get 99% of the vote.

Predictions based on assumptions - especially when there are literally millions of variables (every single voter for one thing) - are generally just that - predictions. I predict I'll win the lotto tomorrow.

Seriously - this is no time to declare certain victory and get complacent. There is a HELL of a lot of work to do before November - and if it isn't done - and done right - that prediction will be, shall we say, less than accurate.

I am reminded of Lieberman saying he has already certainly secured the nomination early in the primary. Such a prediction served him well didn't it?

[TruthIsAll]

14. You fail to see the basic fact that 60-80% of undecided voters

have historically gone to the challenger and against the incumbent. Why should we not expect this to happen this time?

You show a lack of knowledge in predictive modeling. There is no need to introduce the "millions" of variables you are talking about. That is pure hyperbole.

I have used the latest polling data. Of course, the polls are subject to sampling error; that's why I have averaged them and computed the standard deviations- which are a measure of variability and determine an MoE for this poll of polls.

I used 70% as the best guess scenario for Kerry's allocation of the other/undecided vote, taking the mid-point of the historical range. I call that prudent. If you feel its less, then look at the low-point of the range - 60%.

My assumptions are reasonable and unbiased. The statistical methodology is sound.

The caveat that things could change is a given. That's why polls are constantly taken up through election day.

I have given you a snapshot of probabilities based on the best data (most recent polls).

If you disagree with the assumptions or methodology, than present your own analysis.

[Zidane]

16. I call it

a big assumption which is irresponsible. You don't just "assume" 70% will happen in a reliable poll.

Where is your evidence that this assumption will actually happen?

The fact of the matter is that polls that do NOT make assumptions are dead wrong all the time. This is one is less reliable than those due to this assumption.

I'm not saying the prediction couldn't happen, but it should be presented as what it is - a prediction - not a statement of 96% (or whatever) certainty.

My own analysis is this - polls are polls. Ultimately the only poll that will matter is the one taken on election day. For every poll that says one thing there is another one that says just the opposite. No one can responsibly claim that a poll has an upper 90% "certainty" ESPECIALLY when it makes a large "assumption" in order to reach that "certainty". An assumption is NOT a certainty, and to claim something is a certainty when part of the base is generated on a large "assumption" can not leave you with a certainty.

Here is an example. Let's say I have an 80% chance of getting one lotto number correct. Lets also say that I have a 5% chance of getting ALL the lotto numbers correct.

I say "Well, I have an 80% chance of getting one correct. I will just ASSUME that the other numbers will be correct as well. Therefore I have a high probability of winning the lotto." It just isn't logical.

[TruthIsAll]

19. My last response to you on this.

Your lack of knowldege is palpable and obvious.

You:
You don't just "assume" 70% will happen in a reliable poll.
Where is your evidence that this assumption will actually happen?

Me: 60-80% of Undecideds have historically gone to the challenger

You:
The fact of the matter is that polls that do NOT make assumptions are dead wrong all the time. This is one is less reliable than those due to this assumption.

Me: I am making assumptions based on historical fact.


You:
I'm not saying the prediction couldn't happen, but it should be presented as what it is - a prediction - not a statement of 96% (or whatever) certainty.

Me: Oh, your lack of understanding is so obvious. The 96% is not a prediction, it is a probability. You don't know the difference.

You:
My own analysis is this - polls are polls. Ultimately the only poll that will matter is the one taken on election day. For every poll that says one thing there is another one that says just the opposite. No one can responsibly claim that a poll has an upper 90% "certainty" ESPECIALLY when it makes a large "assumption" in order to reach that "certainty". An assumption is NOT a certainty, and to claim something is a certainty when part of the base is generated on a large "assumption" can not leave you with a certainty.

Me:
That's why I look at 16 national polls and 50 state polls.
I have never assumed a certainty; the allocations are all assumptions. That is my prerogative, based on my knowledge of voting patterns.

You:
Here is an example. Let's say I have an 80% chance of getting one lotto number correct. Lets also say that I have a 5% chance of getting ALL the lotto numbers correct.

I say "Well, I have an 80% chance of getting one correct. I will just ASSUME that the other numbers will be correct as well. Therefore I have a high probability of winning the lotto." It just isn't logical.

Me: Have you ever taken a math/stat course? I bet you haven't. If you did, you probably did not do too well.

[Zidane]

21. A superiority complex

usually causes people to be blind to their own mistakes.

You also are using the grade school "I'm too smart for you when you call me on something, so I will just ignore you" line. Not very mature for someone who is so much more intellgent than not only my self, but others who have called you on this.

The fact remains your logic is Some facts + assumption = near certainty. About as accurate as 2 + 2 = 9.

"Have you ever taken a math/stat course?"

Sure have. And I know if my professor had tried to teach that a probability based on an assumption can actually be considered even slightly factual he wouldn't have had his job for very long.

"60-80% of Undecideds have historically gone to the challenger
I am making assumptions based on historical fact"

And the fact remains, by your own admission, you are making assumptions. The fact remains your conclusion is based on one. Assuming that your assumption is correct because it has happened in the past is just another assumption in its self. Again, something that can not honestly be presented as a conclusive statement of fact.

What you SHOULD have done - in order to make this more reliable - would be to actually collect factual numbers from undecideds. But then again I suppose it's a lot easier to just "assume" a variable.

If you did this it actually would enter the world of factual probability rather than "this is my personal belief which is as good as any real number". Try doing this under a professor and see how well it is received.

"The 96% is not a prediction, it is a probability"

Nope - it's a prediction because such a probability can not factually exist when it is arrived at using an assumption.

"I have never assumed a certainty; the allocations are all assumptions"

When you try to pass something off as 96% likely that's what we call a "near certainty". When you arrive at such a conclusion it helps to be based entirely on factual numbers rather than assumptions.

"Have you ever taken a math/stat course? I bet you haven't. If you did, you probably did not do too well"

And I expected you to call me on something I randomly generated in about 5 seconds to make a quick point. The evasion from the real issue is very interesting.

Anyway - there are also too many variables still unaccounted for to make any 96% conclusion. It would be honest to say "this will probably happen" or "it looks like this will happen", but it is not honest to say it is exactly a 96% near certainty because you don't have all the numbers to achieve that kind of accuracy.

[TruthIsAll]

24. You have no clue as to what I am doing.

You have totally misunderstood this simple model, the assumptions, the output.

I suggest you show this to someone who knows something about statistics and forecasting models.

You will learn from the experience.

[Spygame]

20. That is fantastic

[lovuian]

23. I remember the Polls saying Gore Won

and he did win it but Bush stole it away with the help of Congress and Supreme Court

So I don't believe this

Seeing is Believing

[BlueJazz]

25. TruthIsAll...I for one, believe your assessment is quite logical.

It may be true that forces (The Bush Crooks) could, outside of our knowledge, change the outcome shown by your Poll (polls)...but the
fact remains that with the knowledge that is known at the present time, Kerry will win.

If you truly wanted to get silly about the whole concept, you would have to factor in Rain, Asteroids hitting the Earth, Kerry in a Plane crash and many other variables.

Your's the best Crystal ball view I've seen...

[TruthIsAll]

28. Here is a thread rearding undecided voters..

Thanks for the good word.

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=104&topic_id=1972511

[Selwynn]

29. Wake me up when they're swearing him in. Until then, fuck all stats.

Well, not literally "wake me up" heh. I'm awake and working, but basically I'm no more interested in all-cap agitated posts swearing Kerry is almost a sure thing than I am in latest breaking news posts that quote Fox and say Kerry is losing ground.

I just don't care. Polls are viritually pointless.

[Kitsune]

30. This assumes, at the outset, a fair election

Which is something that B*sh will never allow.