A TIA MEA CULPA: THE "SIMPLE" NEP EXPLANATION IS WRONG...

TruthIsAll

(1000+ posts) Send PM | Profile | Ignore

Sun Mar-20-05 11:15 PM
Original message

A TIA MEA CULPA: THE "SIMPLE" NEP EXPLANATION IS WRONG...

Edited on Sun Mar-20-05 11:53 PM by TruthIsAll

Its not that simple. I made a mistake in the spreadsheet which
calculates the votes from the preliminary to the final poll.

BUT THE GUN IS STILL SMOKING. 

THE BUSH TOTALS FOR THE VARIOUS DEMOGRAPHIC GROUPS ADD UP TO
MORE THAN 613 ADDITIONAL RESPONDENTS, WHILE KERRY'S
INCREMENTAL TOTALS GO NEGATIVE. 

I'm surprised no one has caught it yet. These are the correct
numbers. There is no simple 1% incremental adjustment.

But that still leaves us with the last 613 exit poll
respondents who went 100%+ for Bush. Hopefully, the numbers
below will make more sense this time.

I'm sorry for the error.
The Naysayers can breathe easier now.

KP2 = KERRY SHARE IN THE PRELIMINARY EXIT POLL OF 13047
KF = KERRY SHARE OF THE 13660 FINAL 
KChg% = KERRY PERCENT CHANGE FROM PRELIM. TO FINAL 
KPCT = KERRY SUBCATEGORY PERCENTAGE OF 13660
BP2 = BUSH SHARE OF THE 13047 PRELIM.
BF  = BUSH SHARE OF THE FINAL 13660.
BChg% =  BUSH PERCENT CHANGE FROM PRELIM. TO FINAL
BPCT = BUSH SUBCATEGORY PERCENTAGE OF 13660

NOTE THE CHANGES FOR EACH DEMOGRAPHIC.
BUSH IS UP IN EACH, BUT BY VARYING AMOUNTS
  

	KP2	Chg	KF	KChg%	KPct	BP2	Chg	BF	BChg%	BPct
       13047	613	13660			13047	613	13660		

GENDER
Male 	2821	-56	2765	-1.98%	20.24%	3121	335	3456	10.74%	25.30%
Female
	3805	-43	3762	-1.12%	27.54%	3170	370	3541	11.68%	25.92%

EDUCATION
No H.S.	271	2	273	0.67%	2.00%	245	22	268	9.15%	1.96%
H.S. 	1464	-51	1412	-3.51%	10.34%	1378	185	1563	13.42%	11.44%
College
	1901	110	2011	5.78%	14.72%	2063	298	2360	14.43%	17.28%
ColGrad	1628	5	1634	0.34%	11.96%	1696	151	1847	8.89%	13.52%
PostG 	1286	-84	1202	-6.56%	8.80%	887	74	962	8.39%	7.04%

RACE/GENDER
WM	1926	-106	1820	-5.52%	13.32%	2724	325	3049	11.92%	22.32%
WF	2514	-50	2464	-1.98%	18.04%	2782	299	3080	10.74%	22.55%
NWM	900	15	915	1.66%	6.70%	365	44	410	12.18%	3.00%
NWF	1306	26	1332	1.98%	9.75%	373	53	426	14.22%	3.12%


AGE
18-29 	1242	12	1254	0.96%	9.18%	954	91	1045	9.57%	7.65%
30-44 	1726	96	1822	5.57%	13.34%	1761	338	2100	19.20%	15.37%
45-59 	1996	-29	1967	-1.46%	14.40%	1840	250	2090	13.61%	15.30%
60+	1628	-120	1508	-7.38%	11.04%	1730	40	1770	2.33%	12.96%


INCOME
0-$15K 	775	-87	688	-11.16%	5.04%	387	6	393	1.53%	2.88%
15-30 	1155	13	1168	1.15%	8.55%	763	97	861	12.75%	6.30%
30-50 	1493	10	1503	0.67%	11.00%	1349	123	1473	9.15%	10.78%
50-75 	1350	1	1351	0.05%	9.89%	1590	169	1759	10.62%	12.88%
75-100 	831	29	861	3.55%	6.30%	848	204	1052	24.03%	7.70%
100-150646	-15	631	-2.28%	4.62%	761	96	856	12.60%	6.27%
150-200245	-16	229	-6.44%	1.68%	277	40	317	14.58%	2.32%
$200+ 	160	-17	143	-10.62%	1.05%	227	31	258	13.72%	1.89%

IDEOLOGY
Lib	2468	-30	2438	-1.22%	17.85%	344	28	373	8.27%	2.73%
Mod	3347	-27	3319	-0.81%	24.30%	2407	359	2766	14.91%	20.25%
Con	689	8	697	1.13%	5.10%	3531	371	3901	10.50%	28.56%


RELIGION
Prot	2973	-77	2896	-2.61%	21.20%	3872	399	4271	10.31%	31.27%
Cath	1761	-28	1733	-1.58%	12.69%	1726	192	1918	11.11%	14.04%
Jewish	301	2	303	0.62%	2.22%	90	12	102	13.80%	0.75%
Other 	685	23	708	3.30%	5.18%	183	37	220	20.40%	1.61%
None 	913	2	915	0.21%	6.70%	378	45	423	11.92%	3.10%

MILITARY
Yes	1010	-2	1008	-0.17%	7.38%	1292	110	1402	8.51%	10.26%
No	5670	-70	5601	-1.23%	41.00%	4921	567	5489	11.53%	40.18%

WHEN DECIDED
Today 	415	-60	355	-14.40%	2.60%	313	-6	307	-1.85%	2.25%
3Days 	207	93	301	44.87%	2.20%	160	69	229	43.00%	1.68%
Week 	125	6	131	4.70%	0.96%	133	6	139	4.70%	1.02%
Month 	783	-45	738	-5.77%	5.40%	496	105	601	21.23%	4.40%
Before
	5154	-190	4964	-3.68%	36.34%	5154	566	5719	10.98%	41.87%



REGION
East	1665	18	1683	1.09%	12.32%	1177	115	1292	9.81%	9.46%
Midwt	1696	9	1705	0.51%	12.48%	1662	149	1811	8.97%	13.26%
South	1780	56	1836	3.16%	13.44%	2184	351	2535	16.08%	18.56%
West	1452	-86	1366	-5.93%	10.00%	1233	106	1339	8.58%	9.80%


PARTY-ID
Dem 	4462	36	4498	0.81%	32.93%	446	110	556	24.60%	4.07%
Rep 	320	-16	303	-5.13%	2.22%	4201	499	4700	11.88%	34.41%
Ind	1832	-92	1740	-5.00%	12.74%	1585	120	1705	7.54%	12.48%


VOTED IN 2000
NoVote	1264	-10	1254	-0.81%	9.18%	909	136	1045	14.91%	7.65%
Gore	4512	37	4549	0.82%	33.30%	397	109	505	27.43%	3.70%
Bush	481	47	529	9.81%	3.87%	4814	531	5345	11.03%	39.13%
Other	339	-48	291	-14.23%	2.13%	68	18	86	26.85%	0.63%

Printer Friendly | Permalink | | Top

dzika

(1000+ posts) Send PM | Profile | Ignore

Sun Mar-20-05 11:50 PM
Response to Original message

1. Thanks for the update.

Printer Friendly | Permalink | | Top

TruthIsAll

(1000+ posts) Send PM | Profile | Ignore

Mon Mar-21-05 12:35 AM
Response to Reply #1

4. The FOUR RED FLAGS thread has more info - and is correct

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x346284

Printer Friendly | Permalink | | Top

TruthIsAll

(1000+ posts) Send PM | Profile | Ignore

Wed Mar-23-05 09:39 AM
Response to Reply #4

12. MAX% OF 2004 VOTERS WHO VOTED BUSH IN 2000 = 41.26%(IF NONE DIED)

Edited on Wed Mar-23-05 09:39 AM by TruthIsAll

MAX% OF 2004 VOTERS WHO VOTED BUSH IN 2000 = 41.26%(IF NONE DIED)

Preliminary Exit poll of 13047 had 41%. PROBABILITY: NEAR ZERO.
Final Exit poll of 13660 had 43%. Impossible. PROB = ABSOLUTE ZERO

Max % = 50.45 mm Bush 2000 votes /122.6 mm 2004 total votes = 41.26%

Printer Friendly | Permalink | | Top

rdmccur

(622 posts) Send PM | Profile | Ignore

Mon Mar-21-05 12:34 AM
Response to Original message

2. The incremental 1% algorithm

appeared too simple for the tricksters to have implemented.
Probably clever folks them tricksters (pardon my texan).

Printer Friendly | Permalink | | Top

berniew1

(1000+ posts) Send PM | Profile | Ignore

Mon Mar-21-05 12:34 AM
Response to Original message

3. How about putting it all together in a documented "white paper"

Printer Friendly | Permalink | | Top

Carolab

(1000+ posts) Send PM | Profile | Ignore

Mon Mar-21-05 03:41 AM
Response to Original message

5. Does that lend more credibility to the rigged aggregators reports?

Or less?

Printer Friendly | Permalink | | Top

LatePeriduct

(660 posts) Send PM | Profile | Ignore

Mon Mar-21-05 03:59 AM
Response to Reply #5

6. It could lend more or less.

I mean we don't have the audited code in our possession right, that keeps constantly being brought up.

I think this is an excellent point to bring up to the author of the site though, so he can provide a good breakdown.

His email address on yahoo is code_breaker_z to ask questions, address issues.

Printer Friendly | Permalink | | Top

kiwi_expat

(526 posts) Send PM | Profile | Ignore

Mon Mar-21-05 09:21 PM
Response to Original message

7. The 613 extra respondents are not important.

Edited on Mon Mar-21-05 09:31 PM by kiwi_expat

As Skids has pointed out in another thread, Mitofsky switched the Kerry-Bush spread in his final data by RE-weighting the original 13047 respondents.

Please look at the following Ohio precinct totals from NEP's final exit poll data. See how Mitofsky has converted the raw 1092-to-924 Kerry lead to a weighted 1025-to-984 Bush win.

The corresponding respondent records are also available. The individual records show exactly how Mitofsky has weighted the sample data to favour Bush. The re-weighting involves ALL of the respondent records.

Raw Kerry Raw Bush Raw Other Raw No vote KerryWgt BushWgt
12 11 0 2 20.438 19
24 26 0 2 17.743 22.478
12 17 0 0 16.851 26.669
24 11 0 1 26.062 14.481
18 26 0 1 14.486 29.224
17 15 0 0 24.292 21.765
18 31 0 4 13.826 27.383
47 4 0 2 34.588 4.424
31 16 0 0 24.341 14.263
28 1 0 1 38.741 1.36
9 11 0 1 14.795 22.486
33 17 0 1 25.342 13.996
23 29 0 0 18.166 27.103
13 33 0 0 10.313 31.885
18 32 0 0 14.125 29.13
24 23 1 0 18.486 22.018
12 13 0 0 17.459 21.868
11 11 1 1 16.679 20.55
13 16 0 0 15.795 23.998
34 16 0 1 25.029 14.16
40 51 1 0 16.928 26.815
6 9 0 0 13.92 24.626
11 23 0 0 11.021 30.65
46 6 0 1 31.285 4.619
21 10 0 0 22.682 14.445
13 39 0 0 10.792 35.929
8 8 0 0 17.977 19.583
12 16 0 0 17.475 25.475
14 33 0 1 10.924 32.974
19 19 0 0 18.557 21.765
30 14 0 0 24.432 13.53
55 12 0 0 30.291 7.346
22 10 0 0 25.85 13.517
27 26 0 0 20.174 23.441
55 15 0 0 28.332 9.747
33 18 0 0 25.649 16.942
11 34 0 1 9.82 34.471
13 19 0 0 16.479 29.473
28 23 1 0 23.604 20.83
11 18 0 0 14.053 26.169
17 14 0 0 19.772 20.786
35 17 0 1 25.18 14.846
24 24 0 0 17.53 23.887
12 10 0 0 23.108 20.854
24 30 0 1 17.847 24.165
18 25 0 0 14.734 23.625
9 19 0 0 13.355 31.327
22 15 0 0 21.614 17.468
35 8 0 0 33.16 7.574
1092 924 4 22 984.102 1025.12 TOTALS FOR OHIO

These final Ohio exit poll precinct totals are from the Ohio Excel spreadsheets prepared by Blue22. The data, itself, is from the Univ. of Mich. file. (Blue22 has kindly offered to e-mail the Excel spreadsheets to anyone who PMs him their e-mail address.)

Printer Friendly | Permalink | | Top

TruthIsAll

(1000+ posts) Send PM | Profile | Ignore

Mon Mar-21-05 09:56 PM
Response to Reply #7

8. Of course, that is true, but we need some dramatic effect...

Edited on Mon Mar-21-05 10:16 PM by TruthIsAll

The question is why the reweighting of an already plausible result?
Why the re-weighting to an impossible result?

How Voted in 2000
43% Voted for Bush in 2000? Impossible. It gave him 2 million more votes then he actually got, so the 41% preliminary weight was closer to reality. And at least one million of Bush 2000 voters died, an unknown number decided not to vote and others were incapacitated. So the 43% number was off by at least 4%. The odds of this occurrence is 1 in the trillions.

PartyID
the weights changed from 38/35/27 to 37/37/26.
The former mix was constant in each of the last three elections.

Gender
The weights stayed constant at 54/46, but the Kerry/Bush percentages changed. How did Kerry only get 51% of females, down from 54% in the prelim. exit poll? Gore got 54%.

Printer Friendly | Permalink | | Top

kiwi_expat

(526 posts) Send PM | Profile | Ignore

Mon Mar-21-05 10:30 PM
Response to Reply #8

9. Clearly Mitofsky had to switch the spread very fast, to match the results.

Edited on Mon Mar-21-05 10:31 PM by kiwi_expat

I'm sure the short interval he had, to pull the switch, accounts for his lousy job in doing it. I totally agree that his excuses for pulling the switch are quite lame.

Please explain to your readers that they don't need to track down the missing 613 respondents. Someone even claimed that the Univ. of Mich. data was bogus because it (probably) included the 613 extra respondents. :-)

Printer Friendly | Permalink | | Top

LatePeriduct

(660 posts) Send PM | Profile | Ignore

Tue Mar-22-05 03:14 AM
Response to Reply #8

10. Now that's alot of dead people.

I have to say, this reminds me of that movie called "The Sixth Sense" you know.

"I see dead people. Dead people are everywhere. They talk to me."

Lots and lots of dead people still voting for Bush even in the afterlife, now that is just religious miracles!

The author of the blog rigged-aggregators.blogspot.com has figured out those parts, for how that many non-voters could end up in it.

Printer Friendly | Permalink | | Top

kiwi_expat

(526 posts) Send PM | Profile | Ignore

Tue Mar-22-05 10:33 PM
Response to Reply #7

11. Here are 34 of the extra respondents

(Skids has made the identification process so simple, I couldn't resist trying to identify some.)

www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x346440#347706

Printer Friendly | Permalink | | Top

mgr

(616 posts) Send PM | Profile | Ignore

Wed Mar-23-05 05:49 PM
Response to Original message

13. You wanted me to comment on the stunning anomalies, here's one:

Edited on Wed Mar-23-05 05:54 PM by mgr

I cannot find the 13047 here consistently through any of the pre existing categories other than your first row. Or why more respondents know their racial/gender classification better than their gender (row 1 compared to row 3) that voted for Kerry? The delta for each category should also be consistent. I don't have time to address it all, but you have a bit more explaining to do before coming to any conclusions that NEP is anything other than GIGO. It's an anomaly that is easy to hide when using percentage.

Looks like provisionally weighted categories that went a little awry, remember that post you thought was incoherent--it seems to have predicted these outcomes, but with the added benefit of affecting both pre and post re weighing.

By the way, the MOE should be <5% when calculating the Bush 2000 voter total, not 1%. Can't ignore the bias factor now can you (well you can, since you just did).

You never did refute my arguments did you, you have had almost a week.

Asking to address your arguments does not constitute a substantive response, especially in a thread on different topic so others may not see it. I recognize the game--keep the critic bogged down addressing your response, and generate numerous threads so the critic cannot respond adequately to all, so the audience never catches on. It is intellectually dishonest.

Mike

Printer Friendly | Permalink | | Top

DU AdBot (1000+ posts)

Tue Apr 30th 2024, 01:38 PM
Response to Original message

Advertisements [?]

Top

Powered by DCForum+ Version 1.1 Copyright 1997-2002 DCScripts.com
Software has been extensively modified by the DU administrators

Important Notices: By participating on this discussion board, visitors agree to abide by the rules outlined on our Rules page. Messages posted on the Democratic Underground Discussion Forums are the opinions of the individuals who post them, and do not necessarily represent the opinions of Democratic Underground, LLC.