05 December 2012

Preliminary Comparison: Obama and Romney to the polling averages, emphasis on Ohio

On November 6th, 2012, with the Final Electoral Landscape (No. 8), I posted a link to ALL of the polling data, in EXCEL form. Here is that LINK once again.

Though a number of states have yet to certify, it is likely that, outside of New York or perhaps New Jersey, there will be no more major changes in the margins.

Here is a quick look of how Obama did compared to the Polling Average for each state. A positive value means that he outperformed his polling.  A negative value means that he underperformed:

Rank '12 State EV 2012 Margin '12 Swing: '12 / '08 2012 FINAL PA Diff: '12/'12 PA
2 HI 4 42,71 -2,55 27,00 15,71
8 CA 55 23,12 -0,91 15,80 7,32
10 NJ 14 17,64 2,11 11,00 6,64
4 RI 4 27,46 -0,35 21,00 6,46
15 OR 7 12,10 -4,25 6,00 6,10
11 CT 7 17,34 -5,03 11,50 5,84
17 MI 16 9,50 -6,94 4,79 4,71
6 MD 10 26,08 0,64 21,67 4,41
13 ME 4 15,29 -2,03 11,03 4,26
7 MA 11 23,14 -2,67 19,00 4,14
21 IA 6 5,81 -3,72 2,14 3,67
1 DC 3 83,63 -2,30 80,00 3,63
24 CO 9 5,37 -3,58 1,87 3,50
25 VA 13 3,87 -2,43 0,77 3,10
22 NH 4 5,58 -4,03 2,50 3,08
19 WI 10 6,84 -7,06 4,63 2,21
16 NM 5 10,15 -4,98 8,00 2,15
20 NV 6 6,68 -5,81 4,67 2,01
9 DE 3 18,63 -6,35 17,00 1,63
27 FL 29 0,88 -1,93 0,61 1,49
18 MN 10 7,65 -2,59 6,25 1,40
23 PA 20 5,40 -4,91 4,20 1,20
5 NY 29 26,39 -0,47 25,20 1,19
14 WA 12 14,87 -2,21 13,74 1,13
26 OH 18 2,95 -1,63 3,24 -0,29
12 IL 20 16,87 -8,24 17,50 -0,63
3 VT 3 35,60 -1,41 37,00 -1,40

Of the 27 states that Obama won, in 24 of them, he exceeded his polling average. In three of them, he underperformed. Let's take a closer look at the three states where he underperformed.

Illinois and Vermont are now certified, so this data will not change. Both were landslide states. In Illinois, Obama performed just 0.60% under his polling margin average. That is pretty darned accurate. In Vermont, he was 1.40% under his average, which is different than in 2008. First, there was no end poll for Vermont. This Castleton poll was from 08/22/2012, and was pretty much cold coffee. It was also a decidedly higher value than Obama had going in to the election of 2008,where he outperformed his Vermont statistic. So, there is a difference between Illinois and Vermont: in Illinois, we had 2 polls to average, but they were very recent.

Also, on the mirror side of Illinois, there were four states where Obama outperformed his average, between +1.13 and +1.40%: WA, NY, PA and MN. Warning: the NY statistic may change again, so this may become three states instead of four. But for now, this means that there were seven states total within 1.4 points of the actual results, which is pretty damned good.

On to OHIO:

In Ohio, which has yet to be certified and Obama currently stands a +2.95%, the current comparison shows him only 0.29% under his polling average, which is really, really, really damned good. You can't get much closer than that.

Update: Ohio was certified on 12/05/2012 at around 4 pm EST and the end-margin is Obama +2.98%.

Let's take a look at the drivers of that average:

Nr. OHIO Date Sample MoE Obama Romney Und Mar.

AVERAGE (one week): N/A N/A N/A 49,69 46,45 3,86 3,24

AVERAGE (3 days):


49,91 46,71 3,38 3,20

AVERAGE (Final two days):


49,99 46,86 3,15 3,13

AVERAGE (final day):


49,99 46,93 3,09 3,06

Mean Value:


49,89 46,74 3,37 3,16










Most recent (no repeaters, one week):















117 NewsmaxZogby Swing States - Final 06.11.12 832 LV +/-3.5 50,9 43,5 5,6 7,4
116 Angus Reid – Final Poll 05.11.12 572 LV +/-4.1 51 48 1 3
115 Ipsos / Reuters – Final Poll 05.11.12 680 LV +/-4.3 50 46 4 4
114 Gravis (R) - FINAL POLL 05.11.12 1316 LV +/-2.7 50 49 1 1
112 SUSA – Final Poll 05.11.12 803 LV +/-3.5 49 44 7 5
111 Rasmussen – Final Poll 05.11.12 750 LV +/-4.0 49 49 2 0
110 Ohio Poll / U of Cincy 05.11.12 889 LV +/-3.3 50 49 1 1
109 PPP (D) – FINAL POLL 04.11.12 1000 LV +/-3.1 52 47 1 5
108 Pulse (R) / Let Freedom Ring (Tea Party) 04.11.12 1000 LV +/-3.0 49 46 5 3
107 YouGov 04.11.12 1620 LV +/-3.0 49 46 5 3
104 Columbus Dispatch 04.11.12 1501 LV +/-2.2 50 48 2 2
102 Grove Insight (D) / PNA / USAction 03.11.12 500 LV +/-4.4 49 45 6 4
101 NBC / WSJ / Marist 02.11.12 971 LV +/-3.1 51 45 4 6
99 WAA (R) 02.11.12 1649 LV +/-2.6 50,2 45,8 4 4,4
98 CNN/ORC 02.11.12 796 LV +/-3.5 50 47 3 3
95 Wenzel (R) / Citizens United (R) 01.11.12 1281 LV +/-2.7 46 49 5 -3
93 Quinnipiac / NYT / CBS 31.10.12 1100 LV +/-3.0 50 45 5 5
86 Mellmann (D) / AFC 30.10.12

49 44 7 5
85 Pharos Research 30.10.12 765 LV +/-3.5 49 46,3 4,7 2,7

Note that for Ohio, I created four different values, as the Buckeye State, with 117 polls total in 2012, was the most polled state in the Union and with 19 polls in the final 6 days, I kept shortening the window, just to see what would happen. Notice that the value for the final day of polling averages (11/05), was just +3.06%, or just 0.11% off from the current Ohio margin. That is amazingly good.

So, who was on and who was way, way, way off in Ohio?

Well, obviously (surprise, surprise), Zogby, with Obama  +7.4, was way, way, way off to the left, by 4.45 points But Wenzel, which showed (chuckle, chuckle) Romney +3 just 5 days before the voting began, was off by 5.95 points. Put together, those two polls showed a 10.4 variance, something that cannot exist in the same universe at the same time.

But notice that even with Zogby in the final day's mix and with Wenzel in the total 6 day mix, both averages were amazingly close to reality. This once again proves my argument that the average of all polls really does mirror reality, for an individual whackadoodle result to the left pretty much cancels out an individual whackadoodle result to the Right, and so on and so one.

This is the second election cycle in a row where Rasmussen sat on the fence and declared a 49-49 tie for both candidates, and for the second time in a row, Rasmussen was off, this time by 3 points. And though PPP, Quinnipiac and Mellmann, all showing Obama +5, were also off 2 points to the Left, they were less off than Rasmussen was and they all got the winner right.

Right now, the closest pollsters in Ohio were Pharos (Obama +2.7), CNN / YouGov / Angus Reid and Pulse (R), commissioned by "Let Freedom Ring" (Tea Party), all of which showed Obama +3.

This makes for a double problem for Rasmussen, which really has some 'splainin' to do, for you see, Pulse (R), the robo-polling unit, is an owned subsidiary of Rasmussen, who has already admitted that Pulse uses exactly the same technology and the same methodology as Rasmussen himself. Pulse was created to provide a cheaper price to customers. So, how can Pulse (R) and Rasmussen himself produce such divergent results, just one day apart from each other? This is perhaps even a scandal, for one must ask if Ras was deliberately fudging the numbers right before election day, and I have already sent a mail to Rasmussen to this effect.

To their credit, the Tea Party Organization "Let Freedom Ring" published the poll results showing Obama +3. Also, We Ask America (WAA), a rabidly right-wing leaning organization, showed Obama +4.4. Now, around 1 point off in a relatively close race is really understandable. This means that of the limited pool of respondents you had, about 0.5% to 0.7% were actually for the other side, which made a total shift of 1 point or so. So, I have no complaints about the Ipsos/Reuters, the Grove Insight (D), the WAA or the Columbus Dispatch polls, all of which were roughly 1 point off.

I can also maybe understand 2 points off, though I think it means that better polling was necessary, and here is one of the rare cases on election night where PPP (D) was too optimistic in its polling. PPP, Quinnipiac, Mellmann, Gravis (R) and the Ohio Poll were all off by 2 points, either to the Left or to the Right. In a massive landslide state, that would be non-consequential, but in a hotly contested battleground state like Ohio, it is really under par. I am more likely to forgive them in Ohio as the state was also bombarded way too much from all sides in terms of advertising and polling and I can understand it when Ohioans simply started to tune out and turn-off.

But totally unforgivable are the results that were off by 3 or more: Rasmussen, Wenzel, Marist and Zogby. But notice that only one firm (Marist) was really too far off to the Left. The other three were too far off to the Right, with whackadoodle Wenzel (who polls for World Net Daily, which thinks that Obama's wedding ring is secret islamic code and that he had a number of gay lovers, btw) taking the prize for the worst polling of Ohio in 2012.

In terms of Top-line percentages and the margin, Angus Reid, a firm out of Canada, wins the prize for the best polling result in Ohio, for Angus Reid's 51/48, Obama +3 is eerily close to the current Ohio statistic:


DEM GOP IND Total DEM % GOP % IND % Margin Mar % USA
Ohio 2012 2823137 2658469 99538 5581144 50,58% 47,63% 1,78% 164668 2,95% Ohio 2012
Ohio 2008 2940044 2677820 103951 5721815 51,38% 46,80% 1,82% 262224 4,58% Ohio 2008
Diff: -116907 -19351 -4413 -140671 -0,80% 0,83% -0,03% -97556 -1,63% Diff:


Angus Reid has no stake in our politics. It truly is an independent and neutral firm. And looky, looky, a foreign firm did the best polling in the hottest electoral prize of 2012.

In the coming weeks, I will be singling out individual states for analysis just like Ohio, and Conservatives can rest with satisfaction in knowing that I will be just as hard and unforgiving with Democratic leaning firms who really got it wrong in North Carolina, which was a Romney pick-up in 2012. And believe me, there were Democratic firms that screwed up in the Tarheel State.

--------------------------------------------------------------------

How about Romney's margins, compared to the end polling averages?

Rank '12 State EV 2012 Margin '12
2012 FINAL PA Diff: '12/'12 PA
16 AK 3 13,99 -7,55 22,00 -8,01
4 ID 4 31,91 6,61 36,00 -4,09
18 MS 6 11,50 -1,67 15,00 -3,50
6 AR 6 23,69 3,84 27,00 -3,31
15 TX 38 15,79 4,03 17,50 -1,71
20 IN 11 10,20 11,23 11,20 -1,00
21 MO 10 9,38 9,25 10,12 -0,74
23 GA 16 7,82 2,62 8,25 -0,43
2 WY 3 40,82 8,58 40,00 0,82
22 AZ 11 9,04 0,56 7,67 1,37
1 UT 6 48,04 20,02 46,33 1,71
9 KS 6 21,72 6,80 19,80 1,92
12 ND 3 19,62 10,97 17,50 2,12
11 TN 11 20,39 5,33 18,00 2,39
14 LA 8 17,21 -1,42 14,07 3,14
24 NC 15 2,04 2,37 2,40 4,44
19 SC 9 10,47 1,49 6,00 4,47
10 NE* 5 21,92 6,99 17,25 4,67
17 MT 3 13,66 11,28 8,68 4,98
13 SD 3 18,02 9,61 12,00 6,02
8 AL 9 22,19 0,61 15,00 7,19
3 OK 7 33,54 2,25 25,90 7,64
7 KY 8 22,69 6,47 14,00 8,69
5 WV 5 26,76 13,67 14,00 12,76

It is a similar story, but not identical. Of the 24 states that Mitt Romney won, he surpassed his polling averages in 16 of them and fell short in 8 of them, where 3 of those states, which were never polled, were guestimates on my part, but realistic ones. So, in reality, Romney only truly and verifiably underperformed in 5 states: ID, AR, TX, MO and GA. But of those five MO and GA are a deviance of -0.74 and -0.43, respectively, which means that the polling, just as in the cases of IL and OH for Obama, were really, really close.

This is all important data, as the Right cannot possibly claim that all polling was skewed toward Obama. Were this the case, then Romney would not have outperformed the polling averages in 15 states!!!

No, the answer to the question of WHY the polling averages were sometimes really off has to do with the subjects of:

-LV vs. RV models

and

-gross miscalculation of the Latino Vote, something I have been warning about for months on end.

--------------------------------------------------------------------------

And just for fun's sake, a very quick look at Oregon, a state that people kind of forget:

Nr. OREGON Date Sample MoE Obama Romney Und Mar.

AVERAGE N/A N/A N/A 48,75 42,75 8,50 6,00


















9 PPP (D) / LCV – FINAL POLL 02.11.12 921 LV
52 46 9 6
8 DHM / Fox News 12 30.10.12 500 LV +/-4.4 49 42 9 7
7 Elway / Oregonian 30.10.12 505 LV +/-5.0 47 41 9 6
6 Hoffman (R) 29.10.12 615 LV +/-3.42 47 42 9 5
5 SUSA 19.10.12 579 LV +/-3.9 49 42 9 7
4 SUSA 17.09.12 552 LV +/-4.3 50 41 9 9
3 PPP (D) 26.06.12 686 LV +/-3.7 50 42 8 8
2 SUSA 10.05.12 1648 RV +/-2.6 47 43 10 4
1 SUSA 20.03.12 1615 RV +/-2.5 50 39 11 11

Every pollster was off in Oregon, including PPP, but Hoffman (R), which showed 47/42, Obama +5, is hilarious:

"On Wednesday October 24th and Thursday October 25, 2012, The Hoffman Research Group surveyed a cross section of Oregon voters to ascertain voter preferences in the 2012 Presidential election.

The resulting data found President Barack Obama currently leading Governor  Mitt Romney by a 47% to 42% margin.

When pushed, undecided voters broke toward Mitt Romney almost 2 to 1. "

Oh, really? 

Oregon, current results:


DEM GOP IND Total DEM % GOP % IND % Margin Mar %
Oregon 2012 970488 754172 63454 1788114 54,27% 42,18% 3,55% 216316 12,10%
Oregon 2008 1037291 738475 52098 1827864 56,75% 40,40% 2,85% 298816 16,35%
Diff: -66803 15697 11356 -39750 -2,47% 1,78% 0,70% -82500 -4,25%

Gee, Hoffman was only off by 7.1 points...

Now, PPP was also off, by 6.1 points. In both cases, that is way too much to be off, but at least PPP came close to the actual topline results.

And interestingly enough, the very first SUSA poll of Oregon, from MARCH 2012, came the closest to the actual results, but of course was not in the end polling average. Fascinating. In MARCH 2008, the SUSA 50 map that was produced came unbelievably close to the actual results of the election in that year. This is worth more study...

Coming up for comparison fun tomorrow: Connecticut, Michigan and Iowa...

No comments:

Post a Comment

Constructive comments and critique are always welcome. Please keep it polite and respectful.