1996 NATIONAL EXIT POLL - EXACTLY MATCHED RESULTS

[TruthIsAll]

LAST 3 NATIONAL EXIT POLLS:
1996 PARTY ID: 39% Dem / 35% Repub - matched results
2000 PARTY ID: 39% Dem / 34% Repub - matched results
2004 PARTY ID: 38% Dem / 36% Repub - ???????

http://www.cnn.com/ALLPOLITICS/1996/elections/natl.exit.poll/index1.html							

Sample size: 16,359 / Updated: 11/06/96 11:49 a.m.

							
Categ	     Clinton	Dole	Perot
Gender	      48.72%	40.88%	8.44%
Whites 	      43.20%	45.88%	9.44%
Race	      48.76%	41.12%	8.34%
Age	      48.85%	40.53%	8.69%
Income	      48.89%	40.79%	7.45%

PartyId	      48.49%	41.00%	8.47%
1992Vote	48.75%	41.18%	8.16%
Ideology	48.99%	41.14%	8.27%
Education	49.02%	39.64%	9.41%
Religion	48.22%	39.76%	9.23%

w/oPerot 	47.77%	41.10%	7.45%
Decided 	48.63%	39.97%	8.58%
			
Average	      48.15%	41.18%	8.49%
							

	
						
										
			HORIZONTAL		VERTICAL	WEIGHTED TOTAL		
Gender
.......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
Men	48	43	44	10	42	52	57	20.64%	21.12%	4.80%
Women	52	54	38	7	58	48	43	28.08%	19.76%	3.64%
								48.72%	40.88%	8.44%

Whites
.....All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
Men	48	38	49	11	42	51	57	18.24%	23.52%	5.28%
Women	52	48	43	8	58	49	43	24.96%	22.36%	4.16%
								43.20%	45.88%	9.44%
										
Race
......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
Whites	83	43	46	9	73	93	90	35.69%	38.18%	7.47%
Blacks	10	84	12	4	17	3	4	8.40%	1.20%	0.40%
Hispan  5    72	21	6	7	2	3	3.60%	1.05%	0.30%
Asians	1	43	48	8	1	1	1	0.43%	0.48%	0.08%
Other	1	64	21	9	2	1	1	0.64%	0.21%	0.09%
								48.76%	41.12%	8.34%
										
Age
......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
18-29	17	53	34	10	18	14	21	9.01%	5.78%	1.70%
30-44	33	48	41	9	32	33	34	15.84%	13.53%	2.97%
45-59	26	48	41	9	26	27	27	12.48%	10.66%	2.34%
60+    24	48	44	7	23	25	19	11.52%	10.56%	1.68%
								48.85%	40.53%	8.69%
										
Income
.......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
0-$15 	11	59	28	11	14	8	14	6.49%	3.08%	1.21%
$15-30 23	53	36	9	25	20	24	12.19%	8.28%	2.07%
$30-50 27	48	40	10	27	27	31	12.96%	10.80%	2.70%
$50-75	21	47	45	7	20	23	17	9.87%	9.45%	1.47%
$75-100 9	44	48	7	8	10	7	3.96%	4.32%	0.63%
$100k+	9	38	54	6	7	12	7	3.42%	4.86%	0.54%
							      48.89%  40.79%  7.45%
										
PartyID
.......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
Dem	39	84	10	5	68	10	23	32.76%	3.90%	1.95%
Rep	35	13	80	6	9	68	24	4.55%	28.00%	2.10%
Ind	26	43	35	17	23	22	53	11.18%	9.10%	4.42%
							      48.49%	41.0% 8.47%
										
92Vote
......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
Clinton	    43	85	9	4	75	10	22	36.55%	3.87%	1.72%
Bush	    35	13	82	4	10	69	18	4.55%	28.70%	1.40%
Perot	    12	22	44	33	6	13	48	2.64%	5.28%	3.96%
Other	     1	24	36	9	0	1	1	0.24%	0.36%	0.09%
NoVote	9	53	33	11	9	7	11	4.77%	2.97%	0.99%
							      48.75%	41.18%	8.16%
										
Ideology
.......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
Lib	20	78	11	7	31	5	17	15.60%	2.20%	1.40%
Mod	47	57	33	9	55	37	51	26.79%	15.51%	4.23%
Cons	33	20	71	8	14	57	32	6.60%	23.43%	2.64%
							      48.99%  41.14%	8.27%
										
Education
.........All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
No High    6	59	28	11	8	4	8	3.54%	1.68%	0.66%
H.S.Grad  24	51	35	13	25	21	32	12.24%	8.40%	3.12%
College   27	48	40	10	27	27	29	12.96%	10.80%	2.70%
CollGrad  26	44	46	8	23	30	21	11.44%	11.96%	2.08%
PostGrad  17	52	40	5	18	18	10	8.84%	6.80%	0.85%
							     49.02%	39.64%	9.41%
										
Religion
.......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
Protest.  38	41	50	8	32	47	33	15.58%	19.00%	3.04%
Catholic  29	53	37	9	31	27	27	15.37%	10.73%	2.61%
Christian 16	45	41	12	15	17	21	7.20%	6.56%	1.92%
Jewish     3	78    16	3	6	1	1	2.34%	0.48%	0.09%
Other      6	60	23	11	7	3	7	3.60%	1.38%	0.66%
None	    7     59	23	13	9	4	11	4.13%	1.61%	0.91%
							     48.22%	39.76%	9.23%

If Perot did not run, would vote for:
.......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot								
Clinton	50	91	3	5	93	4	30	45.50%	1.50%	2.50%
Dole	43	4	90	5	4	92	30	1.72%	38.70%	2.15%
No Vote	5	11	18	56	1	2	36	0.55%	0.90%	2.80%
								47.77%	41.10%	7.45%
										
When Decided
.......All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
3Days	11	35	38	22	8	11	28	3.85%	4.18%	2.42%
Week	6	35	47	17	4	7	12	2.10%	2.82%	1.02%
Month	13	47	36	13	12	11	19	6.11%	4.68%	1.69%
Before69	53	41	5	75	70	41	36.57%	28.29%	3.45%
							      48.63%	39.97%	8.58%

[helderheid]

1. uscountvotes.org needs you TIA

Please contact Kathy Dopp (who crunched the Florida numbers) at [e-mail] and visit www.uscountvotes.org

You're wonderful and your work is important!

[jkd]

7. Why don't you contact Kathy and have her invite TIA to come on board?

Your requests to TIA don't seem to be working.

[TruthIsAll]

29. CORRECTION: 2004 PARTY ID: 38 Dem /35 Rep /27 Ind

Just to be accurate.

It would be great if they provided the numbers to one decimal
place, like Zogby does. It would seem to make sense, since the
MOE for the National Exit Poll is 1.0%.

It makes sense, so that's why they don't do it.

So we will just use what they give us.

		HORIZONTAL			WEIGHTED		
PARTY ID	MIX	Bush	Kerry	Nader	Bush	Kerry	Nader
Democrat 	38%	9%	90%	1%	3.4%	34.2%	0.4%
Republican 	35%	92%	7%	0%	32.2%	2.5%	0.0%
Independent	27%	45%	52%	2%	12.2%	14.0%	0.5%
	     100%			     47.77%	50.69%	0.92%
				

Probability = 	3.29447E-09	

       or  
          ***** 1 in	303,538,508 ******

That Bush could go from 47.77% to 50.73% in the actual vote.

[IndyPriest]

2. What LUCK! They match!

Because, you know, exit polls weren't built to be able to actually track the votes, just the trends in voters. Amazing luck in '96!

[papau]

3. Seems exit polls get lucky and match when a Bush is not running!

:-)

[Freddie Stubbs]

4. How do we know that numbers weren't changed to match the results

as was done last year? :shrug:

[euler]

8. Exit polls are always reweighted to the actual vote count,

The exit poll wouldn't be of much use to the MSM (who pays for the exit poll) if it's not reweighted to the acual count.

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=284055#285192

[Al-CIAda]

5. TIA, tell the truth...you are the inspiration for the new series

"NUMB3RS" aren't you?

People Lie... Numbers Don't
Inspired by actual cases and experiences, "Numb3rs" depicts the confluence of police work and mathematics in solving crime. The story of an FBI agent who recruits his mathematical genius brother to help solve a wide range of challenging crimes in Los Angeles from a very distinctive perspective.

http://www.tvtome.com/tvtome/servlet/ShowMainServlet/showid-25043/

Just kidding you...great job as always.
:thumbsup:

[flintdem]

6. You're using WEIGHTED results for 1996

Here are the weighted and unweighted resulted for VNS 1996 National Exit Poll (first column shows with missing data and second without missing data).

Weighted:
Clinton 48.3 49.1 N=8037
Dole 40.0 40.6 N=6644
Perot 8.3 8.5 N=1384
Other 1.8 1.8 N=294
Missing 1.6 -- N=271


Unweighted:
Clinton 51.4 52.2 N=8550
Dole 36.9 37.5 N=6133
Perot 8.5 8.6 N=1407
Other 1.7 1.7 N=285
Missing 1.6 -- N=262

It shows a 2.9% skew toward Clinton from the actual election results.


Voter News Service. VOTER NEWS SERVICE GENERAL
ELECTION EXIT POLLS, 1996 .
ICPSR version. New York, NY: Voter News Service
, 1996. Ann Arbor, MI: Inter-
university Consortium for Political and Social
Research , 1997.

[TruthIsAll]

9. And how come these "weighted" results were provided so quickly...

on 11am (the day after the election) and we are still waiting for the 2004 "official" results 3 months later.

[qwghlmian]

11. TIA, do you think you should have mentioned that

you were comparing weighted results of 1996 to raw results of 2004? It is a pretty important distinction, you know.

[rdmccur]

26. Show that these 2004 results

were raw, please. According to the Edison/Mitofsky exit poll data
at 4pm and 7pm these results were 'weighted' (evidenced by the words "time of weighting"). I think the methodology used in 2004 needs to made clear (along with release of the raw data) and it hasn't been to my knowledge. Why?

[qwghlmian]

27. The raw data in 2000, 1996 etc was released

12 months after the election. In 2004, Mitofsky promised (according to MysteryPollster) to release it within 4 months of the election. It will be available from Roper center for a fee, just like the previous elections' raw data. Nothing extraordinary has been happening in 2004, no matter how much you hear otherwise.

As for your post, please explain how the data for 2004 at 4pm and 7pm could possibly be weighted using actual election results, like the data that TIA is quoting from 2000 and 1996 was. The voting was still going on.

[Name removed]

32. Deleted message

Message removed by moderator. Click here to review the message board rules.

[flintdem]

13. Not sure what you mean...

The 1996 data were released in September of 1997. It is usually about 10 months to get the data after the election.


If you are talking aobut the time the weighted data was released on election night...

According to your post the 1996 CNN weighted results were posted at 11:49pm about the same time that the 2004 weighted results were going to be posted until the Mitofsky server went down.

Does this answer your question?

[TruthIsAll]

20. Oh, yes, the Mitofsky server...down just when they needed it

Oh, so it took 10 months to release those numbers.
I see.
It's all very clear now.

[flintdem]

21. Data release dates

I check the release date for presidential exit polls from 1984-2000. August of the following year was the earliest any data was released and one study wasn't released for 14 months.

I checked a little further and found that regular news polls (by abc, cbs, etc.) also took 12-18 months to be released.

[TruthIsAll]

23. I recall seeing an exit poll in the NYTwithin a week of the elections

What are the regular news polls you are referring to?
12-18 months?

This is all very confusing.

[flintdem]

25. I am talking about when the dataset is released...

...sure the marginals are reported (Candidate A 51% Candidate B 47%) right away, but the actual dataset with all of the respondents and all of their individuals answers for all of the questions come much later.

Regular news polls are the polls showing public attitudes on abortion or social security or homeland security- the kind of stuff that gets reported in newspapers and on CNN every week. Those datasets get released into the public domain months after the poll results get published. Abc News does a poll on some issue, including all sorts of demographic questions. They report the main findings on their evening news. A year later they release the entire set of data into the public domain.

I'm starting to wonder...you don't have much familiarity with polling data and the types of datasets used...

[euler]

12. No one here understands the difference between raw numbers...

...collected at the precincts on election day and the actual final exit poll. They terms are used interchangeably in this forum. Previous efforts to point this out have failed. I hope you have better luck.

[flintdem]

10. Here is WEIGHTED & UNWEIGHTED Party ID for 1996

Weighted

Party ID
Frequency % Valid % Cumulative %
Valid Democrat 6070 36.5 39.4 39.4
Republican 5336 32.1 34.7 74.1
Independent 3387 20.4 22.0 96.1
Other 598 3.6 3.9 100.0
Total 15391 92.6 100.0
Missing System 1239 7.4
Total 16630 100.0


Unweighted

Party ID
Frequency % Valid % Cumulative %
Valid Democrat 6495 39.0 42.1 42.1
Republican 4924 29.6 31.9 74.0
Independent 3406 20.5 22.1 96.1
Other 597 3.6 3.9 100.0
Total 15422 92.7 100.0
Missing System 1215 7.3
Total 16637 100.0


Almost a 10% difference between Democrat and gop in the unweighted data.

[TruthIsAll]

14. I have a PartyID split was 39 Dem-35 Rep. Why the disparity?

When were the numbers which you show provided by the media?

And where are the full demographics displayed?
How do they compare to the numbers I found?

[qwghlmian]

15. In the post above yours -

the weighted numbers are 39.4 to 34.7 - that's the 39 to 35.

[TruthIsAll]

16. Hmm.. roundoff. The 4.0% becomes 4.7%. Good. Need to recalc.

Thanks.

PartyId
	All	Clinton	Dole	Perot	Clinton	Dole	Perot	Clinton	Dole	Perot
Dem	39.4	84	10	5	68	10	23	33.10%	3.94%	1.97%
Rep	34.7	13	80	6	9	68	24	4.51%   27.76 2.08%
Ind	25.9	43	35	17	23	22	53	11.14%	9.07%	4.40%
								48.74%	40.77%	8.46%

Removing the roundoff has the effect of raising Clintons
numbers:
His 48.49% before roundoff becomes 48.74% after.

And Dole drops from 41.00% to 40.77%.

[flintdem]

17. Our numbers match...Sloppy columns

Sorry I can't get these columns to stay straight when I post them.

My second column was the percentages with the missing responses included as a category. The third column was the party id without the missing responses. That column weighted shows Dem 39.4 Rep 34.7 and rounding give us the same results 39-35. The last column is the cumulative percentages.

I haven't run all of the demographics. I'm using the actual 1996 national data set in SPSS (with and without the weight variable).

[TruthIsAll]

18. You should use Excel. Much faster. n/t

.

[flintdem]

19. Actually a Stata user, normally :) n/t

[TruthIsAll]

22. Those wonderful, mysterious "weights".

.

[flintdem]

24. Weighting and Polling go hand in hand

"Although pollsters select their samples to representative of the population form which they are drawn, sometimes they must adjust a sample before analyzing and reporting the results of a poll. The adjustments may be made for substantive reasons or because of biases in the characteristics of the selected sample."

"Polling and the Public: What Every Citizen Should Know", Herbert Asher 2004, p. 85

Those of us who used polling data on a daily or weekly basis know that weighting is very common. I would estimate that a third of all dataset I have every used were weighted, and (because I looked at the codebooks)EVERY National exit poll has been weighted.

[TruthIsAll]

28. YES, THEY DO

		HORIZONTAL			WEIGHTED		
PARTY ID	MIX	Bush	Kerry	Nader	Bush	Kerry	Nader
Democrat 	38%	9%	90%	1%	3.4%	34.2%	0.4%
Republican 	35%	92%	7%	0%	32.2%	2.5%	0.0%
Independent	27%	45%	52%	2%	12.2%	14.0%	0.5%
	100%				      47.77%	50.69%	0.92%
				

Probability = 	3.29447E-09	

       or  
          ***** 1 in	303,538,508 ******

That Bush could go from 47.77% to 50.73% in the actual vote.

[ClintCooper2003]

30. Not so fast, TIA. The election that year was on Nov. 5th, not the 6th...

Those figures have obviously been adjusted.

[TruthIsAll]

31. I know Nov 5 is election day. Nov. 6 is the date of the data on the site.

?

[ClintCooper2003]

33. It says updated Nov. 6th - in exactly the same way that CNN adjusted...

figures the following day for the 2004 election.

By the way, I saw an article saying that the exit polls showed Bill Clinton won the 1996 election by about 14-15 points. That is consistent with pre-election polls as well. And yet, he only won by 8.51% on election day. Is it possible that massive fraud has been occuring for quite some time?

[TruthIsAll]

34. One thing is for sure. Millions of Dem votes are spoiled every election.

How were they spoiled?

Greg Palast is the expert on this.