1988-2004 MoE exceeded in 66 of 238 state exit polls (65 for Rep, 1 Dem)

[tiptoe]

1988-2004 MoE exceeded in 66 of 238 state exit polls (65 for Rep, 1 Dem)

TruthIsAll     source: http://richardcharnin.com/StateExitPollDiscrepancies.ht...

Feb. 18, 2010

An analysis of state exit poll discrepancies for the 1988-2004 elections yields an interesting pattern. The data is from Edison Media Research/ Mitofsky International: "Evaluation of Edison/Mitofsky Election System 2004" (January 19, 2005).

E-M provided unadjusted exit poll data for 238 of 255 state presidential elections from 1988 to 2004. They define "Within Precinct Error (WPE)" as the difference between the unadjusted exit poll share margin and the recorded vote-count share margin. “Error” implies that the exit polls were wrong. But millions of votes are uncounted in every election (nearly 11 million in 1988 and 4 million in 2004).  Therefore, it is more accurate to refer to Within Precinct Discrepancy (WPD). A positive WPD indicates that the vote shift favored the GOP; a negative WPD favored the Democrat.

For example, in NY 2004 the 12.2 WPD is the difference between Kerry’s exit poll share margin (30.5%) and recorded vote-count share margin (18.3%). Kerry’s unadjusted exit poll share (64.5%) was 6.1% higher than his recorded share, far exceeding the 2-3% exit poll margin of error (MoE). The WPD is double the share deviation.

 

MoE
 ± 2-3%

Unadjusted 2004 Exit Poll

Share

(a)

Kerry

Bush

Margin

 

Recorded 2004 Vote-Count

Share

(b)

Kerry

Bush

Margin

 

WPD
(a - b)
Margin Δ

 

NY

64.5

34.0

30.5

 

58.4

40.1

18.3

 

12.2

 

Of the 238 state elections — from exit poll to vote-count — 194 shifted to the vote for the Republican and 44 to the Democrat.

 

Shift to
Dem
GOP
Total

2004
5
46
51

2000
14
32
46

1996
10
37
47

1992
4
44
48

1988
11
35
46

Total
44
194
238

Share
18.50%
81.50%
100%

 

One would expect an approximately equal shift. The fact that 82% vote shifted to the Republican over the five presidential elections was either due to errors in the exit polls or vote miscounts. It could not have been due to chance. But we have 238 exit polls, not just a few. Exit polls are known to be quite accurate — outside the USA.

Assuming a 3% MoE for each of the 238 state exit polls:
– In 15 Democratic states, the average WPD was 4.9. The MoE was exceeded in 25 elections, ALL for the Republican.
– In 15 Battleground states, the average WPD was 3.8. The MoE was also exceeded in 25 elections, ALL for the Republican.
– In 21 Republican states, the average WPD was 2.4. The MoE was exceeded in 16 elections, ALL but ONE for the Republican.

Were the discrepancies due to Republican voters reluctance to be polled in each of the five elections? Not likely.
Were they due to Democratic voters misstating how they voted to the exit pollsters and actually voting for the Republican in each of the five elections? Not likely.

Or were they due to the millions of mostly Democratic votes that were uncounted? That is more than likely. It is a fact.
Or were they due to votes that were miscounted in favor of the Republican? That is quite likely.

We assume a conservative 3% unadjusted exit poll margin of error for each state election. Since the vote-share deviation is one-half the WPD, a 3% MoE is exceeded when the WPD is at least 6%. Given the 95% confidence level, approximately 12 of 238 elections would be expected to fall outside the margin of error (6 for the Democrats and 6 for the Republicans). But the MoE was exceeded in 66 elections — 65 in favor of the Republicans. The probability is ZERO that this was due to chance.

The probability that the MoE would be exceeded a given state is 1/20 (5%). It is 1/40 (2.5%) for the Democrat and 2.5% for the Republican.
The probability that the MoE would be exceeded in at least N of 238 state elections is given by the Excel BINOMDIST function:

Prob =1- BINOMDIST (N-1, 238, 2.5%, TRUE)

The following table lists the probability that the MoE would be exceeded in at least N of 238 state exit polls.
The function cuts off at N=34, a 1 in 200 trillion probability.
Imagine what it is for 65 states.

 

N
2
5
10
15
20
25
30
34
---
65

Probability
93.80%
54.80%
3.90%
0.04%
7.90E-07
4.80E-10
1.10E-13
5.00E-15

??????

or 1 in
1
2
26
2,642
1,258,312
2,080,038,070
9,324,222,830,995
200,159,983,438,689

 

Assuming a more realistic 2% exit poll MoE, it would be exceeded if the WPD were at least 4%. The 2% MoE was exceeded in 109 of 238 state elections: 100 in favor of the Republican, 9 in favor of the Democrats.

Let’s take a closer look at the 2004 election.
– The average Democratic state WPD was a whopping 8.9. The 3% MoE was exceeded in 11 of 15 states (73%) — all for Bush.
– The average Battleground state WPD was 6.9. The 3% MoE was exceeded in 10 of 15 states (67%) for Bush and none for Kerry.
– The average Republican state WPD was 3.8. The 3% MoE was exceeded in 7 of 21 states (33%) — all for Bush.

The 3% margin of error was exceeded in a total of 28 states — all in favor of Bush. The probability is 1 in 19 trillion that the MoE would be exceeded in 16 states. Imagine what the probability is for 28 states.

Assuming a 2% MoE, the probability is even lower, since the MoE was exceeded in 36 states: 34 in favor of Bush, 2 in favor of Kerry.

The distribution of the WPD in Democratic, GOP and Battleground states indicates that the GOP strategy was:
1) Pad Bush’s popular vote "mandate" by cutting Democratic margins in heavily populated BLUE states (NY, CA, CT, NJ, MD, MA, MI).
2) Steal the electoral votes in Battleground states (FL, OH, NM, CO, NV, MO, IA).
3) Pad the vote in RED states with large minority (Democratic) voting blocs (TX, MS, AL, TN, SC). Ignore the others (ND, SD, OK, MT, KY).

Unadjusted exit poll data has not been released for 2008. Is it because the data would confirm what the 1988-2004 exit polls indicate? In every election, the Democrats do much better than the official recorded vote indicates. As always, the Final 2008 National Exit Poll was forced to match the recorded vote-count. A True Vote Model (see below) indicates that Obama won by over 22 million votes.

Historical WPD

Within Precinct Discrepancy — Vote Margin  Δ :  State Exit Poll  to  Vote-Count
 
Voting Booth to Exit Poll Interview to (secret) Counting of Candidate & Proposition Votes
 

positive
negative

#
(#)

 ==   %, difference in  vote-share  margins — in favor of the GOP
 ==   %, difference in  vote-share  margins — in favor of the DEM

Combined 50 States+DC

3.82 avg

 

 

WPD,  average  Δ (%)
 

# of STATES with Vote Margin   SHIFTED

from unadjusted Exit Poll to Vote-Count
...in favor of the GOP
...in favor of the DEM

# of STATES with WPD (%)  EXCEEDING
 Exit Poll  Margin of Error  assuming ± 3%


 Exit Poll  Margin of Error  assuming ± 2%

 

2008
 
XX.X   ?

Does the NEP MSM
Dare Release the
Exit Poll Report?

2004

7.40



46
5


28


36

2000

2.01



32
14


4


15

1996

1.93



37
10


7


12

1992

5.40



44
4


20


31

1988

2.38



35
11


7


15

 

5-Election Avg

3.82


5-Election Tot

194
44


66


109

•

81.5

 % of 238 vote-share margin shifts favored the GOP in election vote-counts

 

18.5

 % of 238 vote-share margin shifts favored the DEM in election vote-counts

• 

65

#

(

98.5

 %)  >  ± 3% MoE/XP shifted in favor of GOP in the vote-count

 

1

(#)

(

1.5

 %)  >  ± 3% MoE/XP shifted in favor of DEM in the vote-count

• 

100

#

(

91.7

 %)  >  ± 2% MoE/XP shifted in favor of GOP in the vote-count

 

9

(#)

(

8.3

 %)  >  ± 2% MoE/XP shifted in favor of DEM in the vote-count

 

15 Democratic States

4.9 avg

 

 

WPD,  average  Δ (%)
 

# of STATES with WPD (%)  EXCEEDING
Exit Poll  Margin of Error  assuming ± 3%


Exit Poll  Margin of Error  assuming ± 2%

 

CA
CT
DE
DC
HI

IL
ME
MD
MA
MI

NJ
NY
RI
VT
WA

 

2004

 8.9 


11


13

11.6
16.0
15.9
2.8
8.2

3.5
4.0
7.3
7.7
6.4

9.1
12.2
5.3
15.2
8.0

2000

3.0


2


4

3.8
0.9
7.1
na
na

6.4
2.1
4.3
4.3
2.2

0.4
3.3
0.4
(0.4)
3.7

1996

1.8


0


4

4.7
(4.2)
1.3
na
na

(1.2)
4.4
3.6
3.3
3.5

1.9
(2.1)
1.1
5.0
2.0

1992

7.3


9


13

8.5
8.3
7.3
na
na

6.1
5.1
8.1
7.1
4.9

11.2
4.6
9.0
8.6
5.5

1988

4.0


3


8

3.8
5.2
1.8
na
0.6

2.6
6.4
5.2
6.8
1.2

4.2
7.2
(0.2)
5.8
5.4

 

5-Election Avg

4.9

5-Election Tot

25


42

6.5
5.2
6.7
2.8
4.4

3.5
4.4
5.7
5.8
3.6

5.4
5.0
3.1
6.8
4.9

• 

25

#

(

100.0

 %)  >  ± 3% MoE/XP shifted in favor of GOP in the vote-count

 

0

(#)

(

0.0

 %)  >  ± 3% MoE/XP shifted in favor of DEM in the vote-count

• 

41

#

(

97.6

 %)  >  ± 2% MoE/XP shifted in favor of GOP in the vote-count

 

1

(#)

(

2.4

 %)  >  ± 2% MoE/XP shifted in favor of DEM in the vote-count

 

15 Battleground States

3.8 avg

 

 

WPD,  average  Δ (%)
 

# of STATES with WPD (%)  EXCEEDING
Exit Poll  Margin of Error  assuming ± 3%


Exit Poll  Margin of Error  assuming ± 2%

 

CO
FL
IA
MN
MO

NV
NH
NM
NC
OH

OR
PA
VA
WV
WI

 

2004

 6.9 


10


13

6.1
7.8
3.0
9.2
5.8

9.9
14.0
8.0
11.9
10.6

1.8
8.4
8.7
(5.9)
4.8

2000

0.8


2


5

5.6
0.6
(3.0)
(0.5)
(1.8)

6.0
2.4
(5.1)
9.8
1.0

na
0.8
2.0
(4.5)
(2.4)

1996

3.8


4


5

1.5
0.6
(0.3)
(1.7)
5.8

na
12.2
7.0
6.5
3.1

2.4
3.6
6.5
2.6
2.8

1992

5.2


6


9

7.2
5.5
2.0
6.4
8.6

(2.1)
10.1
6.3
4.2
4.4

13.6
2.0
3.5
3.2
2.5

1988

1.9


3


3

3.0
2.4
0.6
(1.2)
0.6

3.0
6.0
6.6
0.4
1.6

7.4
0.8
(0.6)
0.2
(2.2)

 

5-Election Avg

3.8

5-Election Tot

25


35

4.7
3.4
0.5
2.4
3.8

4.2
8.9
4.6
6.6
4.1

6.3
3.1
4.0
(0.9)
1.1

• 

25

#

(

100.0

 %)  >  ± 3% MoE/XP shifted in favor of GOP in the vote-count

 

0

(#)

(

0.0

 %)  >  ± 3% MoE/XP shifted in favor of DEM in the vote-count

• 

32

#

(

91.4

 %)  >  ± 2% MoE/XP shifted in favor of GOP in the vote-count

 

3

(#)

(

8.6

 %)  >  ± 2% MoE/XP shifted in favor of DEM in the vote-count

[Wilms]

1. Those darn levers are the fly in your ointment. Ain't they?

Since you can't flip votes on lever machines, and their undervote rates have been OK, there's no WAY you can claim logic behind your assertion regarding NY State.

I actually BELIEVE the election was stolen (unlike you who is damn sure). Your illogical idea of lever voting, and your refusal to get yourself educated about it tells me your acting as an advocate, not as a scientist or statistician. Which is ok. It just ain't compelling.

[Bill Bored]

2. The Toe has already conceded that lever machines can't switch votes.

I don't have time to find that particular post but it's around.

But s/he has never answered the question of how many machines it would take to account for exit poll discrepancies using unintentional undervotes, or how that could physically be accomplished. The short answers are:
LOTS of machines and CAN'T be done, respectively.

[tiptoe]

3. Fly in the ointment?

NY votes were CAST on levers

...levers with "sensor latches" deactivated can and have for decades caused undervotes, mostly minority.

They were COUNTED by central tabulators.

Those are computers

...as are proprietary-coded, programmable DRE & op-scan recording-and/or-secret-counting/re- voting machines.

An election-win strategy of the 'hijacked GOP' -- from which lifelong "traditional GOP" Susan Eisenhower recently de-registered, along with a reported 344,000 other disgusted 'true GOP', from data drawn from just 28 states since 2006 -- is revealed in words by loose-lipped NY-Rep Peter King remindful of the paraphrased reputed wisdom attributed to this democracy-loving thug:

It's not who votes that counts.

It's who (and what) does the counting.

...which, insinuated, shouldn't be expected to necessarily match a true vote.

In summer 2003 in the presence of *, the NY Rep asserted with no uncertainty the upcoming '04 national election result, explaining not in terms of typical political bravado such as "we will get out the vote" or "we will nail the debates" or "we will have the voters' support" (only 48.5% * approval on election day and never a King 2003 concern) but in terms suggesting as strongly as the Rep's own confidence ("It's already over...the election's over, we won...it's all over...") that any preparations necessary for * reelection 16 months down the road had been completed and were already in place by mid-2003, and all that remained was -- not the voting per se on election day but rather -- "...the counting..." and even elaborating: "...we'll take care of the counting". Note he didn't even take care to be discreet: He did not say 'it's all over but the voting' or even 'the counting of the vote'...merely 'the counting'. Seemingly by 2003 "counting" had become euphemism for "voting" with the 'hijacked GOP'.

The '04 NY state WPD itself shows huge discrepancy between NY exit poll share margin (30.5% w  ± 3.21% MoE, cluster adjusted) and the recorded NY vote-count share (18.3%). The problem involves more than preserving the presence and use of a certain vote-recording device. Levers, even with sensor latches activated (and presuming such status is sustained through elections), are limited in number in a context of ever-growing electorate long-lines and experience other mechanical "lockup" problems.

(links: "secret", "Counting", "Votes" and "GOP" in Historical WPD section)

Still unexplained, too, is the breadth of discrepancy between exit poll share margin and recorded vote-count share margin (after allowing for exit poll MoE with cluster effect).

You have been advocating for NY to keep the levers, claiming that "votes cannot be switched" on mechanical levers as they can be on touch screens or optical scanners.

That is a misleading assertion and a half-truth.

There are three possibilities:
1) Perhaps you have been UNAWARE all along that although all NY votes were CAST on levers, the precinct totals were MANUALLY INPUT to central COMPUTERS to TABULATE the TOTAL STATE VOTE.

2) Perhaps you have been AWARE that NY computers tabulated the votes but assumed they are NOT VULNERABLE TO MALICIOUS CODING AND/OR HACKING - unlike the Diebold/ES&S computers in other states that ARE vulnerable.

3) Perhaps you have been AWARE that NY computers tabulated the votes and that they ARE VULNERABLE TO HACKING AND/OR MALICIOUS CODING like the Diebold/ES$S computers in other states. But you chose NOT to reveal this in your posts.

So which is it: 1,2 or 3?

If it is 1)then you were just ill-informed.

If it is 2) can you PROVE that the PRECINCT VOTE TOTALS entered into the COMPUTERS were a) accurate and b) that the software did not switch Kerry votes to Bush?

If it is 3)...well, let others draw their own conclusions.

Perhaps you are also unaware that in 2004 CT was the only other state besides NY to vote exclusively on levers and that the 16.0 WPD was the second highest in the nation. It's in the table below.

Why were the exit pollsters so far off the mark in NY and CT — the only 100% LEVER states?

You say that you believe Kerry won. Therefore you must also believe that the exit polls were essentially correct - except, that is, for NY and CT.
Agenda much?

[Wilms]

4. There was NO significant undervote rate.

So your sensor argument is pointless.

More drivel about King. It doesn't mean the lever were tampered with. And if the election was stolen, it didn't have to involve NY or CT.

As for the tabulation, have you looked at precinct data? And if the precincts were hand counted how do you suppose the aggregation would be done? More to the point, what does the tabulation have to do with the lever? And if the police mainframe, or the news service screws up the numbers (which is pretty much assured) what of the recanvass. You ALWAYS skip that. We'll let Kellner explain: http://www.democraticunderground.com/discuss/duboard.ph...

And of course you don't see the illogic behind your assumption that my assumption that Kerry won makes me a believer in ANYONE's cooked numbers.

Five years in the kitchen.

[OnTheOtherHand]

5. could you ramble more?

I find it all too easy to follow your points.

I don't know whether TIA supposes that all the lever machines were somehow connected to GEMS through the Intertubes, or what.

[Bill Bored]

6. That sensor latch was reinstalled prior to 2004 so...

...it's not hard to measure the effect that it had on the undervote rate in prior years. Does it account for the Exit Poll discrepancies? Doubt it.

Every voting system uses some form of "central tabulator." Even hand counts. Otherwise, there would have to be one gigantic precinct for a whole state.

Hacking a central tabulator would be about the dumbest way to try to rig an election because it's very easy to check these tallies against the original election-night counts, be they machine counts or hand counts. It's possible that really dumb party or candidate representatives could miss something, but not very likely because both sides want to win their elections. The kind of systemic fraud you are suggesting to account for the exit poll discrepancies would be readily detectable during the 100% recanvass of every precinct in the state of NY.

What motivated this latest round of out-of-the-woodwork lever bashing anyway? Some trouble in ballot scanner paradise?

PS -- The clever way to rig an election is to switch the votes as they are cast, at the precinct, on election day, as we know occurred in Cuyaghoga County, Ohio using paper ballots (punch cards in that case), ballot order rotation within the poll sites, DREs, etc., programmed to do so BEFORE the election on EMSs, as in Erie County, NY in 2009 (probably accidentally), and which is NOT POSSIBLE with lever voting machines!!!