The Final Pre-election LV and RV Polls Do Not Match

Why the Final Pre-election Polls and the National Exit Poll Confirm an Obama 20m Vote Landslide

Jan. 25, 2009

This analysis will show why the final likely voter (LV) and registered voter (RV) polls, in conjunction with the National Exit Poll, confirm that Obama won a landslide at least double his 9.5m official vote margin. The mathematical confirmation is not obvious and requires a focus on the structure (and anomalies) in the pre-election and exit polls.

Adjustments must be applied to the final pre-election polls and National Exit Poll to derive an approximation to the True Vote. The following facts should be clearly understood so that the adjustments make sense intuitively.

- RV polls sample registered voters.

- LV polls sample likely but not new registered voters. They are a subset of RV polls.

- Obama’s average 13% RV poll margin was nearly double his LV poll margin.

- From 1992-2004, the Democrats won new voters by a 14% average margin.

- In 2008, there were approximately 20.8m new voters (131.37m - 110.6m returning 2004 voters).

- According to the Final National Exit Poll (NEP) Obama had 71% (14.7m) of 20.8 million new voters.

- Final pre-election polls indicated that approximately 5% of voters were undecided.

- Undecided voters typically break 70-90% for the challenger (Obama).

- Based on the final LV 9-poll average with undecided voters allocated, Obama had 53.96% (59.7 of 110.6m returning voters).

- Combining new and returning voters, Obama won by 74.4 - 54.9m (56.7 - 41.8%)

- Obama had an average 56.96% share in the final 3 RV polls - within 0.3% of the combined LV (returning voter) and new voter share.

- The Final National Exit Poll (NEP) is always “forced to match” the official recorded vote count.

- In 2004, the Final NEP Bush/Gore 43 / 37% voter mix implied there were 52.6m Bush 2000 voters in 2004; he had 50.5m in 2000.

- In 2006, the Bush/Kerry returning voter mix was an implausible 49 / 43%. The Democratic landslide was denied.

- In 2008, the impossible Bush/Kerry 46 / 37% mix implied there were 11m more returning Bush than Kerry voters; Bush won by 3m votes.

- The Election Calculator model confirmed both the LV poll / new voter share and the RV projection

In order to determine the number of returning 2004 and new 2008 voters, voter mortality and turnout rate estimates were applied to the recorded 2004 vote. The calculation indicated that there were approximately 110.6m returning voters. Since 131.37m votes were recorded, there were approximately 20.77m new voters (15.8% of the total). Based on the 71-27% NEP share, Obama won new voters by 14.7-5.6m.

Pollsters usually allocate 60-90% of the undecided vote (UVA) to the challenger. The UVA is a function of the incumbent's job performance. Obama’s UVA was projected at 75%. This was plausible since he was the "challenger". McCain represented an incumbent with a 22% approval rating.

Considering that the LV polls do not include new voters, why not break the projection of the total vote into two parts? Let’s use the average LV poll to calculate the Obama/McCain shares of returning (likely) voters and use the National Exit Poll to calculate their share of new voters.

According to the National Exit Poll (NEP), 13% (17m) of the 2008 electorate consisted of new voters and returning third-party voters 4% (5.2m). But 4% is impossible because only 1.2m third-party votes were recorded in 2004. If the third-party percentage is changed to the true 1%, then applying 3% to new voters raises the percentage to 16%, which means that there were 21m new voters, not 17m as the NEP indicates.

Although this does not have a material effect on the total Obama share (he had 71% of new voters and 66% of other) it does indicate that more weight should be given to pre-election RV polls which included new voters than to the LV polls which did not.

We can approximate the number of returning 2004 voters by applying the 1.2% annual U.S. voter mortality rate (4.8% since 2004) and assuming that 95% of living 2004 voters returned to vote in 2008 and 4 % were returning third-party voters (5.2m). But 4% is impossible because only 1.2m third-party votes were recorded in 2004. If the third-party percentage is changed to the true 1%, then applying 3% to new voters raises the percentage to 16% (21m) new voters, not 17m as the NEP indicates. Although this does not have a material effect on the total Obama share (he had 71% of new voters and 66% of other) it does indicate that more weight should be given to the pre-election RV polls, which included new voters, than the LV polls which did not.

In 2004, 122.3m votes were recorded. Applying the 4.8% mortality rate and 95% turnout, 110.6m returned to vote in 2008. Since 131.37m votes were recorded, simple subtraction shows that there were approximately 20.8, new voters.

The National Exit Poll was forced to match the recorded 52.87-45.62% recorded vote – but the returning voter mix was impossible (see below).

The 2008 Election Calculator confirmed the above analysis:

Obama had 55.3%, assuming the NEP “Voted in 2004” shares and the 2004 recorded vote (adjusted for mortality and turnout).
Obama had 57.5%, assuming the NEP “Voted in 2004” shares and the 2004 unadjusted exit poll (adjusted for mortality and turnout).

Based on the above logic, this confirms the accuracy of the RV and LV polls:

1.The pre-election LV polls were correct in predicting the likely voter (returning) vote shares.

2.The pre-election RV polls were correct in predicting the registered (returning and new) vote shares.

3. The LV polls by themselves are misleading; they understated the true Obama margin by 6%.

4. The RV polls, adjusted for undecided voters, were accurate.

5. Obama won the True Vote with a 57% share with better than a 20 million vote margin.

Why did RCP fail to list these 5 RV polls?

Poll Date Obama McCain Spread

Gallup      11/02          53        40           13
ABC/WP   11/02         54        41           13
Pew         11/01          50        39           11
Pew         10/26          52        36           16
NBC          10/20          52        42           10

Total 52.2 39.6 12.6

LV and RV Projections of Returning 2004 Voters

2008 Total Obama McCain Other

Recorded	131.37	69.46	59.93	1.98
Share	100%	52.87%	45.62%	1.51%

Assumptions

75% UVA to Obama

2004 Voters
4.8% mortality (5.87m)
95% turnout of lving 2004 voters in 2008 (110.61m)

2008 National Exit Poll share of new voters
Obama 71%
McCain 27%
Other 2%

Calculating the 2008 True Vote

(Assume final LV polls for returning voters and National Exit Poll for new voter shares)

Final November Pre-election Polls

Poll	Obama	McCain	Other	Undecided
LV (9)	51.00%	43.56%	1.50%	3.94%
RV (3)	52.33%	40.00%	1.50%	6.17%

Allocate undecided voters

UVA

Obama

UVA 75%

McCain

25% UVA

Other

53.96%

44.55%

1.50%

56.96%

41.55%

1.50%

Calculate returning voters from prior election

Total 2004	122.29
Died	5.87	4.8%
Alive	116.42
Turnout	110.60	95.0%	(LV)

Total 08	131.37
New	20.77

Vote	Obama	McCain	Other	Total
NEP	71%	27%	2%	100%
New	14.75	5.61	0.42	20.77

Returning	53.96%	44.55%	1.50%	100.0%
Turnout	59.67	49.27	1.66	110.60

True Vote

Total	74.42	54.87	2.07	131.37
Share	56.65%	41.77%	1.58%	100.00%

Recorded Vote

Total	69.46	59.93	1.98	131.37
Share	52.87%	45.62%	1.51%	100.0%

Final RV Polls (11/02)

Poll	Poll	Sample	Obama	McCain	Spread	Obama 75% UVA	McCain 25% UVA	Spread
10RV	Date	Average	52.33	40.00	12.33	56.96	41.55	15.42
Gallup	10/31 - 11/02	2847RV	53	40	13	57.125	41.375	15.75
ABC/WP	10/31 - 11/02	2446RV	54	41	13	56.625	41.875	14.75
Pew	10/29 - 11/01	1325RV	50	39	11	57.125	41.375	15.75

LV and RV Poll Averages (10/03-11/03)

Average	Obama	McCain	Spread	Obama 75% UVA	McCain 25% UVA	Spread
10 RV	50.30	40.00	10.30	56.45	42.05	14.40
25 LV	50.84	42.60	8.24	54.64	43.87	10.77

Final 10 RV Polls (10/04-11/02)

Poll	Poll	Sample	Obama	McCain	Spread	Obama 75% UVA	McCain 75% UVA	Spread
10 RV	Date	Average	50.30	40.00	10.30	56.45	42.05	14.40
Gallup	10/31 - 11/02	2847RV	53	40	13	57.13	41.38	15.75
ABC/WP	10/31 - 11/02	2446RV	54	41	13	56.63	41.88	14.75
Pew	10/29 - 11/01	1325RV	50	39	11	57.13	41.38	15.75
Pew	10/23 - 10/26	1325RV	52	36	16	59.88	38.63	21.25
NBC/WSJ	10/18 - 10/20	1159RV	52	42	10	55.38	43.13	12.25

Ipsos	10/09 - 10/13	1036RV	48	39	9	56.63	41.88	14.75
Newsweek	10/08 - 10/09	1035RV	52	41	11	56.13	42.38	13.75
FOX News	10/08 - 10/09	900RV	46	39	7	56.13	42.38	13.75
Ipsos	10/02 - 10/06	858RV	47	40	7	55.63	42.88	12.75
NBC/WSJ	10/04 - 10/05	658RV	49	43	6	53.88	44.63	9.25

Final 25 LV Polls (10/03-11/03)

	Poll	Poll	Sample	Obama	McCain	Spread	Obama 75% UVA	McCain 25% UVA	Spread
	25 LV	Date	Average	50.84	42.60	8.24	54.64	43.87	10.77
1	Research2k	10/31 - 11/02	1100LV	51	44	7	53.63	44.88	8.75
2	Zogby	11/01 - 11/03	1201LV	54	43	11	55.13	43.38	11.75
3	Hotline/FD	10/31 - 11/02	882LV	50	45	5	52.63	45.88	6.75
4	Rasmussen	10/31 - 11/02	3000LV	51	46	5	52.13	46.38	5.75
5	Battleground	10/27 - 10/30	1000LV	49	45	4	52.38	46.13	6.25

6	Marist	11/03 - 11/03	804LV	52	43	9	54.63	43.88	10.75
7	FOX News	11/02 - 11/03	971LV	50	43	7	54.13	44.38	9.75
8	NBC/WSJ	11/01 - 11/02	1011LV	51	43	8	54.38	44.13	10.25
9	CNN	10/30 - 11/01	1017LV	51	43	8	54.38	44.13	10.25
10	Pew	10/30 - 11/01	2587LV	49	42	7	54.63	43.88	10.75

11	CBS	10/28 - 10/31	1005LV	54	41	13	56.63	41.88	14.75
12	Marist	10/29 - 10/29	543LV	50	43	7	54.13	44.38	9.75
13	FOX News	10/28 - 10/29	924LV	47	44	3	52.63	45.88	6.75
14	Ipsos	10/23 - 10/27	831LV	50	45	5	52.63	45.88	6.75
15	Newsweek	10/22 - 10/23	882LV	53	41	12	56.38	42.13	14.25

16	CBS/NYT	10/19 - 10/22	771LV	52	39	13	57.63	40.88	16.75
17	FOX News	10/20 - 10/21	936LV	49	40	9	56.13	42.38	13.75
18	Pew Resrch	10/16 - 10/19	2382LV	53	39	14	57.88	40.63	17.25
19	Ipsos	10/16 - 10/20	773LV	50	42	8	54.88	43.63	11.25
20	CNN	10/17 - 10/19	764LV	51	46	5	52.13	46.38	5.75

21	Pew Resrch	10/12 - 10/14	1191LV	50	40	10	56.38	42.13	14.25
22	CBS/NYT	10/10 - 10/13	699LV	53	39	14	57.88	40.63	17.25
23	LAT	10/10 - 10/13	1030LV	50	41	9	55.63	42.88	12.75
24	ABC/WP	10/09 - 10/11	766LV	53	43	10	54.88	43.63	11.25
25	CBS/NYT	10/03 - 10/05	616LV	48	45	3	52.13	46.38	5.75

Probability Obama Exceeds Given Vote Share for various MoE

(based on 11/2 RV Polls, 5293 Sample MoE = 1.33%)

2-party	Obama Share >	1.00%	1.20%	1.33%	1.50%	1.75%	2.00%
			Probability
55%	54.18%	100%	100%	100%	100%	100%	100%
56%	55.16%	100%	100%	99%	99%	97%	96%
57%	56.15%	93%	89%	86%	83%	80%	77%
57.74%	56.87%	50%	50%	50%	50%	50%	50%
58%	57.13%	31%	34%	35%	37%	39%	40%
59%	58.12%	1%	2%	3%	5%	8%	11%
60%	59.10%	0%	0%	0%	0%	1%	1%

2008 National Exit Poll

(forced to match the 2008 recorded vote, but the returning voter mix is impossible)

2004	Mix	Votes	Obama	McCain	Other	Obama	McCain	Other
New	13.00%	17.1	71%	27%	2%	12.13	4.60	0.34
Kerry	37.17%	48.8	89%	9%	2%	43.51	4.35	0.98
Bush	45.83%	60.2	17%	82%	1%	10.34	49.27	0.60
Other	4.00%	5.3	66%	24%	10%	3.48	1.25	0.53

Total	100.0%	131.37m	52.87%	45.28%	1.86%	69.46m	59.48m	2.45m

Election Calculator: Scenario 1

(returning voter mix based on the recorded 2004 vote (Bush 50.73-Kerry 48.27%) adjusted for 4.8% voter mortality and 95% turnout)

2004	Mix	Votes	Obama	McCain	Other	Obama	McCain	Other
New	15.80%	20.8	71%	27%	2%	14.83	5.51	0.42
Kerry	40.64%	53.4	89%	9%	2%	47.52	4.80	1.07
Bush	42.71%	56.1	17%	82%	1%	9.54	46.01	0.56
Other	0.85%	1.1	66%	24%	10%	0.74	0.26	0.11

Total	100.00%	131.37m	55.28%	43.08%	1.64%	72.62m	56.59m	2.16m

Election Calculator: Scenario 2

(returning voter mix based on the 2004 exit poll (Kerry 52-Bush 47%) adjusted for 4.8% voter mortality and 95% turnout)

2004	Mix	Votes	Obama	McCain	Other	Obama	McCain	Other
New	15.80%	20.8	71%	27%	2%	14.83	5.51	0.42
Kerry	43.78%	57.5	89%	9%	2%	51.19	5.18	1.15
Bush	39.57%	52.0	17%	82%	1%	8.84	42.63	0.52
Other	0.85%	1.1	66%	24%	10%	0.74	0.26	0.11

Total	100.00%	131.37m	57.54%	40.78%	1.67%	75.59m	53.58m	2.20m