Pollster Report Card: Swing State Edition

I've been working on this post all weekend, and I hope we can have a little fun with it. One of the favorite pastimes of the media is to go back after a presidential election and see which pollsters were closest to the actual result and which ones fell short. Generally, such amateur analyses are the not fairest way to appraise hard-working pollsters, but everyone loves to rank things (especially me), even if it is an inexact science. For his part, Nate Silver has his own way of computing pollster rankings.

Why pollster ranking is tricky, and our swing state focus

The reason it is difficult, and perhaps unfair to rate pollsters post-election is because you pretty much have to go off on the pollster's final product closest to the election as the complete body of work for that firm. The reason for this is simple: a poll for a November race taken in May is unverifiable. Sure, such a poll may have been entirely accurate at the time it was taken, but this is impossible to know because the race is not until November. Therefore, using such an early poll before an election, or even averaging that result with later results is useless in making a fair assessment of that pollster's work. Furthermore, even if we could know that a May poll is right doesn't mean anything: we want to know how close the pollster was to the ultimate result.

With this criterion in mind, I am going to try to do something a little different in ranking the pollsters. Initially, I intended on going through all 50 states. However, I realized that one, this would take way too long, and two, how well a pollster did in states that were uncontested really does not interest me. As a result, I am going to go through all of the swing states and see which pollsters had the best records. (Eventually I will get to national polls.) I want to look at swing state polling because this is where pollsters' work product was most valuable, and also where we will see the most polls being done. Nothing against places like Tennessee or Massachusetts, but because these were not swing states, there was less polling done on them and the results of these states were never in doubt.

Here are the 17 states I decided to take a look at: Arizona, Colorado, Florida, Georgia, Indiana, Michigan, Minnesota, Missouri, Montana, Nevada, New Hampshire, New Mexico, North Carolina, North Dakota, Ohio, Pennsylvania, and Virginia.

In analyzing the work of the pollsters of these states, we will keep it to all polls taken closest to the election, and not earlier than late-September. We will take the last poll a particular pollster did in case it polled the state several times (which is generally the case with the top firms). I realize that using one poll to rate a pollster may not seem entirely, but I think it is quite the contrary. The last poll gives the pollster the best chance to gauge the electorate closest to the election. Furthermore, at the end, we will take each pollster's accuracy for all of his combined work to render a judgment.

Our methodology and the T2L Pollster Score

Naturally, the evaluation process is highly subjective, but I think I have a fair process in place to give us an idea of pollster strength. Here is our methodology: Each poll will get a score based on how many points it was within each of the two candidates' final percentages in a given state combined. We will call this rating a "T2L Pollster Score." At the end, we will average the score for each of the pollster's individual state polls and then list the averages from best to worst.

In other words, say Rasmussen's last New Jersey poll finds the state at 50-48 with Obama winning, and ultimately Obama wins the Garden State 53-46. Consequently, Rasmussen would get a score of 5.0 points on that poll because it was three points off of Obama's final percentage, and two points off of McCain's. So, the lower the score a pollster gets, the better it did. I will eventually take every pollsters' state score and tabulate an average for the pollster (depending on how many swing states it polled). At the end, we list all the firms in order of average closeness in the 17 swing states.

I realize that the "science" here is inexact because most pollsters round up or down their findings, while we are using the actual results to a tenth of a percent. This may be a bit unfair to pollsters, but there is nothing I can do since that is the number we generally get (minus crosstabs) from just about every poll we see. Besides, because every pollster is being judged the same way, no one is being hurt or helped any more than the next one.

I will put up a subsequent summary post going over the rankings at the bottom. Additionally, when we get to the final rankings, I will separate into two categories, the pollsters who did multiple states, and those that generally focused solely on one state. Of one minor note, to save space (and also cut down on confusion), I am just using the pollsters' names, and not the company/network/newspaper which funded the poll (so, if a poll is a Fox/Rasmussen, I am referring to it solely as Rasmussen, but also providing all links).)

Anyway, onto the festivities!


Arizona (Actual Result: McCain 53.7, Obama 45.1)

Politmetrix (10/18-11/01): McCain 50, Obama 46 (4.6)
R2K (10/28-10/30): McCain 48, Obama 47 (7.6)
ARG (10/28-10/30): McCain 50, Obama 46 (4.6)
Mason-Dixon (10/27-10/28): McCain 48, Obama 44 (6.6)
CNN/Time (10/23-10/28): McCain 53, Obama 46 (1.6)
N. Ariz Univ (10/18-10/28): McCain 49, Obama 41 (8.8)
Rasmussen (10/26): McCain 51, Obama 46 (3.6)
ASU (10/23-10/26): McCain 46, Obama 44 (8.8)
Myers/Grove (10/23-10/24): McCain 48, Obama 44 (6.8)
Zimmerman (10/16-10/19): McCain 44, Obama 42 (12.8)

I realize Arizona was not really a swing state, but because the state seemingly tightened in the final weeks, there was a lot of polling done there at the end, and as a result, I thought it was worthwhile to see if any of the pollsters correctly saw that the state was not likely to be close.

On the whole, it looks like the pollsters had some trouble with Arizona. The state, which got on everyone's radar near the end of October, attracted a lot of new polls, only one of which was right-on. The problem was not Obama's number -- 45% -- which eight of the pollsters were within two points of. Rather, just about everyone underestimated McCain's near 54%. The reasons for this might be that pollsters undersampled Republicans (a distinct possibility which R2K did), or perhaps a majority of late-deciders broke to the homestate favorite McCain. Because around ten percent or less decided in the last week, the exit polls do not provide their data for these voters because the sample was so small. Pity.

Colorado (Actual Result: Obama 53.5, McCain 44.9)

Rasmussen (11/02): Obama 51, McCain 47 (4.6)
Polimetrix (10/18-11/01): Obama 55, McCain 40 (6.4)
PPP (10/28-10/30): Obama 54, McCain 44 (1.4)
ARG (10/28-10/30): Obama 52, McCain 45 (1.6)
Mason-Dixon (10/28-10/29): Obama 49, McCain 44 (5.4)
Marist (10/27-10/28): Obama 51, McCain 45 (2.6)
CNN/Time (10/23-10/28): Obama 53, McCain 45 (0.6)
Natl Journal (10/23-10/27): Obama 48, McCain 44 (6.4)
Insider Advantage (10/26): Obama 53, McCain 45 (0.6)
AP/GfK (10/22-10/26): Obama 50, McCain 41 (7.4)
Rocky Mountain/News 4 (10/21-10/23): Obama 52, McCain 40 (6.4)
Zogby Internet (10/17-10/20): Obama 48, McCain 48 (8.6)
Suffolk (10/10-10/13): Obama 47, McCain 43 (8.4)
Quinnipiac (10/8-10/12): Obama 52, McCain 43 (3.4)

As you can see above, we have a wealth of polling in Colorado, a good portion of which did a suburb job.

Florida (Actual Result: Obama 50.9, McCain 48.4)

Survey USA (10/31-11/3): Obama 50, McCain 47 (2.3)
Zogby (10/31-11/3): Obama 49, McCain 48 (2.3)
Rasmussen (11/2): McCain 50, Obama 49 (3.5)
Datamar (11/1-11/2): McCain 48, Obama 47 (4.3)
PPP (10/31-11/2): Obama 50, McCain 48 (1.3)
Strategic Vision (10/31-11/2): Obama 49, McCain 47 (3.3)
Quinnipiac (10/27-11/2): Obama 47, McCain 45 (7.3)
Polimetrix (10/18-11/1): Obama 49, McCain 47 (3.3)
ARG (10/29-10/31): Obama 50, McCain 46 (3.3)
Mason-Dixon (10/28-10/29): Obama 47, McCain 45 (7.3)
Florida COC (10/26-10/27): Obama 45, McCain 41 (13.3)
CNN/Time (10/23-10/28): Obama 51, McCain 47 (1.5)
LA Times/Bloomberg (10/24-10/27): Obama 50, McCain 43 (6.3)
Natl Journal (10/23-10/27): Obama 45, McCain 44 (10.3)
Suffolk (10/23-10/26): Obama 49, McCain 44 (6.3)
AP/Gfk (10/22-10/26): Obama 45, McCain 43 (11.3)
Insider Advantage (10/22): Obama 48, McCain 47 (4.3)
Herald/News 9 (10/20-10/22): Obama 49, McCain 42 (8.3)
Zogby Internet (10/17-10/20): Obama 49, McCain 45 (5.3)
Times Union/Sun Sentinel (10/13-10/15): Obama 49, McCain 45 (5.3)
Hamilton (10/10-10/15): Obama 47, McCain 43 (9.3)

Wow. There was a lot of polling done in Florida this year. Note that these are just polls beginning after October 10. One thing that interests me is that of the polls we have sample/crosstab data for, the pollsters seemed to have a hard time figuring out the right party mix. Many of them seemingly over-sampled both parties, though especially Democrats, while badly undersampling independents and unaffiliated. For many of them, this did not impact the accuracy of their findings, but it is certainly salient.

I think that some pollsters underestimated the impact of many conservatives identifying themselves as indies rather than with the unpopular GOP brand. This is a theme I want to look at a separate post, by the way. Also, the over-sampling of Democrats stands out by itself, and I think that at least in Florida, this state was not as clear a victory for Obama as places like Colorado or Nevada, for example. That we had such a wide variance of results, with some bad findings from top-notch polling names says something about the difficulty of striking the right balance in the Sunshine State this year.

Georgia (Actual Result: McCain 52.4, Obama 47)

Insider Advantage (11/2): McCain 48, Obama 47 (4.4)
PPP (10/31-11/2): McCain 50, Obama 48 (3.4)
Strategic Vision (10/31-11/2): McCain 50, Obama 46 (3.4)
S-USA (10/30-11/2): McCain 52, Obama 45 (2.4)
Politmetrix (10/18-11/1): McCain 50, Obama 47 (2.4)
Rasmussen (10/30): McCain 52, Obama 47 (0.4)
R2K (10/28-10/30): McCain 47, Obama 44 (8.4)
CNN/Time (10/23-10/28): McCain 52, Obama 47 (0.4)
Mason-Dixon (10/22-10/23): McCain 49, Obama 43 (7.4)
Democracy Corps (10/16-10/19): McCain 46, Obama 44 (9.4)

As SUSA, PPP, and R2K demonstrate, Georgia presented some polling problems because it was hard to say exactly what black, and in turn Democratic turnout would be in the end. PPP took the risk, so to speak, and went with very high black turnout, and they ended up correct. However, PPP also overestimated Democratic turnout a bit more than the other two (and presumably other pollsters).

When the voters were tallied, it turns out that three percent more Dems showed up to vote, but more self-identified indies as well. Again, I personally think that pollsters underestimated the number of conservatives who would call themselves indies rather than Republicans (here as well as in other states. Whether this made a big difference in how they voted (and in turn, whether polling was right) is hard to say.

Indiana (Actual Result: Obama 49.9, McCain 49)

Zogby (10/31-11/3): McCain 50, Obama 45 (5.9)
PPP (10/31-11/2): Obama 49, McCain 48 (1.9)
Polimetrix (10/18-11/1): MCain 53, Obama 45 (8.9)
ARG (10/29-10/31): Obama 48, McCain 48 (2.9)
S-USA (10/27-10/30) Obama 47, McCain 47 (4.9)
Rasmussen (10/28-10/29): McCain 49, Obama 46 (3.9)
Selzer (10/26-10/28): Obama 46, McCain 45 (7.9)
R2K (10/24-10/28): Obama 47, McCain 47 (4.9)
Howey-Gauge (10/23-10/24): McCain 47, Obama 45 (6.9)
Big 10 Battleground (10/19-10/22): Obama 51, McCain 42 (8.1)
CNN/Time (10/3-10/6): McCain 51, Obama 46 (5.9)

This was one of the toughest contests to hit spot-on because it was just so close. I predicted that John McCain would squeeze a victory along the lines of 51-to-49, and some very good pollsters agreed. Ultimately, I was incorrect.

Michigan (Actual Result: Obama 57.4, McCain 40.9)

Mitchell Research (11/3): Obama 54, McCain 38 (6.3)
Polimetrix (10/18-11/1): Obama 55, McCain 42 (3.5)
EPIC-MRA (10/28-10/31): Obama 53, McCain 37 (8.3)
PPP (10/28-10/30): Obama 55, McCain 42 (3.5)
Strategic Vision (10/27-10/29): Obama 54, McCain 41 (3.5)
Rasmussen (10/28): Obama 53, McCain 43 (6.5)
Detroit News (10/26-10/28): Obama 50, McCain 38 (10.3)
Big 10 Battleground (10/19-10/22): Obama 58, McCain 36 (5.5)
Quinnipiac (10/8-10/12): Obama 54, McCain 38 (6.3)

Because Michigan was a blow-out, we have less polling to scrutinize for October, and some of the heavy hitters stayed out. Polling a blow-out is in many ways tougher than gauging a tough race. When a contest gets out of hand, it is harder to say who might come our or stay home, and this makes things volatile. I would wager that if we were to take a look at the polling in places like Arkansas, Tennessee, Alabama, and others, besides seeing little polling to look at, we would find wildly varying results.

Minnesota (Actual Result: Obama 54.2, McCain 44)

Survey USA (10/30-11/1): Obama 49, McCain 46 (7.2)
Polimetrix (10/18-11/1): Obama 51, McCain 45 (4.2)
Star Tribune (10/29-10/31): Obama 53, McCain 42 (3.2)
Rasmussen (10/28): Obama 55, McCain 43 (1.8)
PPP (10/28-10/30): Obama 57, McCain 41 (5.8)
R2K (10/27-10/29): Obama 53, McCain 38 (7.2)
Mason-Dixon (10/27-10/28): Obama 48, McCain 40 (10.2)
Minn. Public Radio (10/24-10/28): Obama 56, McCain 37 (8.8)
Big 10 Battleground (10/19-10/22): Obama 57, McCain 38 (8.8)
St. Cloud St. Univ (10/14-10/22): Obama 42, McCain 37 (19.2)
Natl Journal (10/16-10/20): Obama 50, McCain 40 (8.2)
Quinnipiac (10/8-10/12): Obama 51, McCain 40 (7.2)
ARG (10/4-10/7): Obama 48, McCain 46 (8.2)
CNN/Time (9/28-9/30): Obama 54, McCain 43 (1.2)

Like Michigan, Minnesota was a blow-out state in the Electoral College, and it was for basically all of October. This made polling of the state hard enough, but Minnesota's quirky voters made pollsters' jobs here even trickier. S-USA's troubles in the state exemplify some of these problems.

Missouri (Actual Result: McCain 49.5, Obama 49.3)

Rasmussen (11/02): McCain 49, Obama 49 (0.8)
Zogby (10/31-11/3): McCain 49, Obama 49 (0.8)
S-USA (10/30-11/1) McCain 48, Obama 48 (2.8)
PPP (10/31-11/2): McCain 49, Obama 49 (0.8)
Mason-Dixon (10/29-10/30): McCain 47, Obama 46 (5.8)
Insider Advantage (10/29): McCain 50, Obama 47 (2.8)
ARG (10/28-10/30): McCain 48, Obama 48 (2.8)
CNN/Time (10/23-10/28): McCain 50, Obama 48 (1.8)
R2K (10/20-10/23): Obama 48, McCain 47 (3.8)
Suffolk (10/17-10/19): McCain 45, Obama 44 (9.8)
Zogby Internet (10/17-10/20): McCain 48, Obama 48 (2.8)

Along with Indiana and North Carolina, Missouri was the tightest state contest in the country, and the late-breaking polls from there both tell that story and show that several pollsters were at their best in the Show-Me State.

Montana (Actual Result: McCain 49.7, Obama 47.2)

PPP (10/31-11/2): Obama 48, McCain 47 (3.5)
Polimetrix (10/18-11/1): McCain 54, Obama 44 (7.5)
ARG (10/28-10/30): McCain 49, Obama 46 (1.9)
R2K (10/28-10/30): McCain 48, Obama 44 (4.9)
Rasmussen (10/29): McCain 50, Obama 46 (1.5)
Mason-Dixon (10/23-10/25): McCain 48, Obama 44 (4.9)
Montana St. Univ. (10/16-10/20): Obama 44, McCain 40 (12.9)

As you can see, there was not much polling done in Montana, even as it appeared that the state was getting tighter in the closing weeks. This state ended up very close, and Obama almost won it, but fell short. I picked it as one of my two upset specials in the presidential. D'oh!

Nevada (Actual Result: Obama 55.1, McCain 42.7)

ARG (10/31-11/3): Obama 52, McCain 47 (7.4)
Zogby (10/31-11/3): Obama 53, McCain 42 (2.8)
PPP (10/31-11/2): Obama 51, McCain 47 (8.4)
Polimetrix (10/18-11/1): Obama 51, McCain 46 (7.4)
Mason-Dixon (10/28-10/29): Obama 47, McCain 43 (8.4)
Rasmussen (10/27): Obama 50, McCain 46 (8.4)
R2K (10/25-10/28): Obama 50, MCain 45 (7.4)
Suffolk (10/26): Obama 50, McCain 40 (7.8)
CNN/Time (10/23-10/28): Obama 52, McCain 45 (5.4)
AP/Gfk (10/22-10/26): Obama 52, McCain 40 (5.8)
Zogby Internet (10/17-10/20): McCain 52, Obama 44 (20.6)
Insider Advantage (10/19): Obama 47, McCain 47 (12.4)

Nevada is a state that was considered a toss-up probably until the final weeks, when it became apparent because of polling and early voting statistics in Clark and Washoe counties that Obama was pulling away.

So, why were pollsters so wildly off in the Silver State? Well, both CNN's and ABC's exit polls agree on the following splits for the state's voting electorate: 48% male/52% female; 69% white/10% black/15% Hispanic; 38% Dem/30% GOP/32% Independent or other.

Of the data I have access to, PPP found splits of 46 M/54 F; 70 W/8 B/16 H; and 44 D/41R/14I. ARG had the party split in its sample at 43 D/40R/17I. While this represents only two polls, both PPP and ARG heavily oversampled Republicans. And why not? Nevada is a libertarian-leaning state, and it had only voted Democratic twice in national races (the Clinton elections) since 1948. It has been a GOP-rich state for some time. I think this fact played heavily with some pollsters who were not convinced that the state's electorate was going to award Obama a landslide victory. A close win is one thing, and that is what many of them saw happening, but a huge win, even with the Democratic-favorable early voting numbers may have been difficult for many firms to accept.

New Hampshire (Obama 54.4, McCain 44.7)

UNH (10/30-11/2): Obama 53, McCain 42 (4.1)
Polimetrix (10/18-11/1): Obama 54, McCain 40 (5.1)
Rasmussen (10/30): Obama 51, McCain 44 (4.1)
S-USA (10/29-10/30): Obama 53, McCain 42 (4.1)
R2K (10/28-10/30): Obama 51, McCain 44 (4.1)
ARG (10/28-10/30): Obama 56, McCain 41 (5.3)
Suffolk (10/27-10/29): Obama 53, McCain 40 (6.1)
Strategic Vision (10/27-10/29): Obama 50, McCain 41 (8.1)
AP/Gfk (10/22-10/26): Obama 55, McCain 37 (8.3)
Mason-Dixon (10/23-10/25): Obama 50, McCain 39 (10.1)
Marist (10/22-10/23): Obama 50, McCain 45 (4.7)
Zogby Internet (10/9-10/13): Obama 48, McCain 43 (8.1)
CNN/Time (10/3-10/6): Obama 53, McCain 45 (1.7)

New Hampshire also seems to have given the pollsters a bit of trouble. Like Nevada, this may have to do with initial wisdom that New Hampshire was a swing state when in the end Obama pulled away and turned that early narrative on its head. Not sure, though, and I need to look more closely pollsters' available crosstabs.

New Mexico (Actual Result: Obama 56.7, McCain 42)

Polimetrix (10/18-11/1): Obama 53, McCain 43 (4.7)
Survey USA (10/29-10/31): Obama 52, McCain 45 (7.7)
Albuquerque Journal (10/28-10/30): Obama 51, McCain 43 (6.7)
PPP (10/28-10/30): Obama 58, McCain 41 (2.3)
Rasmussen (10/28): Obama 54, McCain 44 (4.7)
*Rasmussen (10/13): Obama 55, McCain 42 (1.7)
Zogby Internet (10/17-10/30): Obama 46, McCain 46 (14.7)

Frankly, I was very surprised there was so little polling in the Land of Enchantment in the final month. I realize that the state got out of hand even before there, but there was plenty of polling in places in New Hampshire, and even Iowa since they became blow-outs.

Anyway, we have very few polls to look at here. You will note that I included two Rasmussen polls. I did this to show that its mid-October one was really close to the actual split, but the one two weeks later -- Rasmussen's last -- made the race even tighter for some reason. So, I wanted to give Ras some extra credit here (though I will not include that earlier poll in Rasmussen's final averages when I tabulate my the T2L Pollster Scoring averages).

North Carolina (Actual Results: Obama 49.9, McCain 49.5)

Rasmussen (11/2): McCain 50, Obama 49 (1.4)
Zogby (10/31-11/3): McCain 50, Obama 49 (1.4)
ARG (10/31-11/3): Obama 49, McCain 48 (2.4)
S-USA (10/30-11/2): McCain 49, Obama 48 (2.4)
PPP (10/31-11/2): Obama 50, McCain 49 (0.6)
Polimetrix (10/18-11/1): Obama 51, McCain 47 (3.6)
Mason-Dixon (10/29-10/30): McCain 49, Obama 46 (4.4)
R2K (10/28-10/30): Obama 47, McCain 45 (7.4)
Insider Advantage (10/29): Obama 48, McCain 48 (3.4)
Civitas (10/27-10/29): Obama 47, McCain 46 (6.4)
CNN/Time (10/23-10/28): Obama 52, McCain 46 (5.6)
Natl Journal (10/23-10/27): Obama 47, McCain 43 (9.4)
AP/GfK (10/22-10/26): Obama 48, McCain 46 (5.4)
WSOC-TV (10/20-10/21): Obama 48, McCain 46 (5.4)
Zogby Internet (10/17-10/20): Obama 50, McCain 47 (2.6)

Like with razor-close Missouri, several pollsters brought their A-game to North Carolina. Note that the poll closest to the actual result came from Public Policy Polling, which is based in North Carolina. I guess PPP knows how to poll its state.

North Dakota (Actual Result: McCain 53.3, Obama 44.7)

Polimetrix (10/18-11/1): McCain 50, Obama 43 (5)
R2K (10/28-10/29): McCain 47, Obama 46 (7.6)
North Dakota UTU (10/13-10/14): Obama 44, McCain 41 (13)
MSU/The Forum (10/6-10/8): Obama 45, McCain 43 (10.6)
*R2K (9/16-9/17): McCain 53, Obama 40 (5)

As you can see, there was not much polling done in North Dakota, and what little there was turned out to be way off. So much for my second presidential swing state upset pick. Ugh. Note that I included two R2K polls to show that one of R2K's earlier versions turned out much closer than its last one. Like with that Rasmussen poll in New Mexico, I won't include it when I tabulate our T2L scores, but I wanted to show that R2K had something on the ball here.

Ohio (Actual Result: Obama 51.2, McCain 47.2)

Rasmussen (11/2): Obama 49, McCain 49 (4)
Zogby (10/31-11/3): Obama 49, McCain 47 (2.4)
S-USA (10/30-11/2): Obama 49, McCain 47 (2.4)
Strategic Vision (10/31-11/2): McCain 48, Obama 46 (6)
PPP (10/31-11/2): Obama 50, McCain 48 (2)
Univ. of Cin. (10/29-11/2): Obama 52, McCain 46 (2)
Quinnipiac (10/27-11/2): Obama 50, McCain 43 (5.4)
Polimetrix (10/18-11/1): Obama 51, McCain 45 (2.4)
Mason-Dixon (10/29-10/30): McCain 47, Obama 45 (6.4)
Columbus Dispatch (10/22-10/31): Obama 52, McCain 46 (2)
LA Times/Bloomberg (10/25-10/27): Obama 49, McCain 40 (9.4)
CNN/Time (10/23-10/28): Obama 51, McCain 47 (0.4)
Natl Journal (10/23-10/27): Obama 48, McCain 41 (9.4)
Marist (10/24-10/26): Obama 48, McCain 45 (5.4)
AP/GfK (10/22-10/26): Obama 48, McCain 41 (9.4)
Insider Advantage (10/22): Obama 52, McCain 42 (6)
Ohio Newspaper Poll (10/18-10/22): Obama 49, McCain 46 (3.4)
Big 10 Battleground (10/19-10/22): Obama 53, McCain 41 (8)
Zogby Internet (10/17-10/20): McCain 49, Obama 47 (2.4)
Suffolk (10/17-10/19): Obama 51, McCain 42 (5.4)
ARG (10/4-10/7): Obama 48, McCain 45 (5.4)

A lot of America's top pollsters are going to be able to send their kids to college with the money made in the field in Ohio this year. As you can see, the results were all over the map, even in the last few weeks.

Pennsylvania (Actual Result: Obama 54.7, McCain 44.3)

S-USA (10/31-11/3): Obama 52, McCain 43 (4)
Zogby (10/31-11/3): Obama 51, McCain 41 (7)
Rasmussen (11/1): Obama 52, McCain 46 (4.4)
Strategic Vision (10/31-11/2): Obama 51, McCain 44 (4)
PPP (10/31-11/2): Obama 53, McCain 45 (2.4)
Polimetrix (10/18-11/2): Obama 51, McCain 44 (4)
ARG (10/29-10/31): Obama 51, McCain 45 (4.4)
Quinnipiac (10/27-11/2): Obama 52, McCain 42 (5)
Mason-Dixon (10/27-10/28): Obama 47, McCain 43 (9)
Marist (10/26-10/27): Obama 55, McCain 41 (3.6)
Insider Advantage (10/26): Obama 51, McCain 42 (6)
Muhlenberg (10/24-10/28): Obama 53, McCain 42 (4)
CNN/Time (10/23-10/28): Obama 55, McCain 43 (1.6)
AP/GfK (10/22-10/26): Obama 52, Obama 40 (7)
Franklin & Marshall (10/21-10/26): Obama 53, McCain 40 (6)
Temple Univ. (10/20-10/26): Obama 50, McCain 41 (8)
Big 10 Battleground (10/19-10/22): Obama 51, McCain 41 (7)
Natl Journal (10/16-10/20): Obama 51, McCain 41 (7)
Susquehanna (10/16-10/18): Obama 48, McCain 40 (11)
Zogby Internet (10/9-10/13): Obama 52, McCain 40 (7)

It kind of makes me laugh how the McCain pushed all-in Pennsylvania, and the pollsters followed, even though the state was never really that close. Interestingly, for the most part, few pollsters hit the eventual 10+ point win for Obama. Perhaps several firms saw this one tightening simply because of McCain's heavy presence in the state. We will go back and look at the (available) PA crosstabs eventually.

Virginia (Actual Result: Obama 52.3, McCain 46.8)

Rasmussen (11/2): Obama 51, McCain 47 (1.5)
Zogby (10/31-11/3): Obama 52, McCain 45 (2.1)
ARG (10/31-11/3): Obama 51, McCain 47 (1.5)
PPP (10/31-11/2): Obama 52, McCain 46 (1.1)
S-USA (10/30-11/1): Obama 50, McCain 46 (3.1)
Polimetrix (10/18-11/1): Obama 52, McCain 45 (2.1)
Mason-Dixon (10/29-10/30): Obama 47, McCain 44 (8.1)
Marist (10/26-10/27): Obama 51, McCain 47 (1.5)
CNN/Time (10/23-10/28): Obama 53, McCain 44 (3.5)
Natl Journal (10/23-10/27): Obama 48, McCain 44 (7.1)
AP/GfK (10/22-10/26): Obama 49, McCain 42 (8.1)
Wash. Post (10/22-10/25: Obama 52, McCain 44 (3.1)
Roanoke College (10/19-10/26): Obama 48, McCain 39 (12.1)
VCU (10/20-10/22): Obama 51, McCain 40 (8.1)
Zogby Internet (10/17-10/20): Obama 50, Obama 46 (3.1)
CNU Virginia (10/11-10/14): Obama 53, McCain 47 (0.9)
Suffolk (10/3-10/5): Obama 51, McCain 39 (9.1)

For the most part, the Virginia pollsters did a really solid job. For a state that had not voted Democratic in a presidential election since 1964, and one that was very tight for most of the election, the pollsters here made a strong showing (with some exceptions, of course).

T2L's Scorecard

Now for the moment you have all been waiting for, where we average out all of the pollsters' scores and determine which ones had the best work in this year swing states.

The following is a list of every single pollster who had a swing state poll in the last four to seven weeks, along with the number of polls, where they were taken, and each pollster's T2L Scores.

Again, the score provided is a number which represents on average, how close that pollster was to the two candidates' final percentages each, combined. The numbers with each state represent the score that pollster got for that particular state (given its' individual outcome).


(1) Opinion Research (CNN/Time). Score: 2.4
(13 states polled: AZ (1.6), CO (0.6), FL (1.5), GA (0.4), IN (5.9), MN (1.2), MO (1.8), NV (5.4), NH (1.7), NC (5.6), OH (0.4), PA (1.6), VA (3.5)).

(2 ) Public Policy Polling. Score: 2.74
(14 states polled: CO (1.4), FL (1.3), GA (3.4), IN (1.9), MI (3.5), MN (5.8), MO (0.8), MT (3.5), NV (8.4), NM (2.3), NC (0.6), OH (2), PA (2.4), VA (1.1)).

(3) Zogby. Score: 3.09
(8 states polled: FL (2.3), IN (5.9), MO (0.8), NV (2.8), NC (1.4), OH (2.4), PA (7), VA (2.1)).

(4) Rasmussen. Score: 3.44
(16 states polled: AZ (3.6), CO (4.6), FL (3.5), GA (0.4), IN (3.9), MI (6.5), MN (1.8), MO (0.8), MT (1.5), NV (8.4), NH (4.1), NM (4.7), NC (1.4), OH (4), PA (4.4), VA (1.5)).

(5) Marist. Score: 3.56

(5 states polled: CO (2.6), NH (4.7), OH (5.4), PA (3.6), VA (1.5)).

(6) Survey USA. Score: 3.94
(11 states polled: FL (2.3), GA (2.4), IN (4.9), MN (7.2), MO (2.8), NH (4.1), NM (7.7), NC (2.4), OH (2.4), PA (4), VA (3.1)).

(7) ARG. Score: 3.98
(13 states polled: AZ (4.6), CO (1.6), FL (3.3), IN (2.9), MN (8.2), MO (2.8), MT (1.9), NV (7.4), NH (5.3), NC (2.4), OH (5.4), PA (4.4), VA (1.5)).

(8) Polimetrix. Score: 4.69
(16 states polled: AZ (4.6), CO (6.4), FL (3.3), GA (2.4), IN (8.9), MI (3.5), MN (4.2), MT (7.5), NV (7.4), NH (5.1), NM (4.7), NC (3.6), ND (5), OH (2.4), PA (4), VA (2.1)).

(9) Strategic Vision. Score: 5.38
(5 states polled: FL (3.3), MI (3.5), NH (8.1), OH (6), PA (4)).

(10) Insider Advantage. Score: 5.61
(7 states polled: FL (4.3), GA (4.4), MO (2.8), NV (12.4), NC (3.4), OH (6), PA (6)).

(11) Quinnipiac. Score: 5.77
(6 states polled: CO (3.4), FL (7.3), MI (6.3), MN (7.2), OH (5.4), PA (5)).

(12) R2K. Score: 6.33
(10 states polled: AZ (7.6), GA (8.4), IN (4.9), MN (7.2), MO (3.8), MT (4.9), NV (7.4), NH (4.1), NC (7.4), ND (7.6)).

(13) Mason-Dixon. Score: 7.23
(13 states polled: AZ (6.6), CO (5.4), FL (7.3), GA (7.4), MN (10.2), MO (5.8), MT (4.9), NV (8.4), NH (10.1), NC (4.4), OH (6.4), PA (9), VA (8.1)).

(14) Big 10 Battleground. Score: 7.48
(5 states polled IN (8.1), MI (5.5), MN (8.8), OH (8), PA (7).

(15) Zogby Internet. Score: 7.52
(10 states polled: CO (8.6), FL (5.3), MO (2.8), NV (20.6), NH (8.1), NM (14.7), NC (2.6), OH (2.4), PA (7), VA (3.1)).

(16) Suffolk. Score: 7.56
(7 states polled: CO (8.4), FL (6.3), MO (9.8), NV (7.8), NH (6.1), OH (5.4), VA (9.1)).

(17) AP/GfK Score: 7.84
(8 states polled: CO (7.4), FL (11.3), NV (5.8), NH (8.3), NC (5.4), OH (9.4), PA (7), VA (8.1)).

(18) LA Times/Bloomberg Score: 7.85
(2 states polled: FL (6.3), OH (9.4)).

(19) National Journal Score: 8.26
(7 states polled: CO (6.4), FL (10.3), MN (8.2), NC (9.4), OH (9.4), PA (7), VA (7.1)).


The following were state polls which had only one poll. I could include these with the rest of the ones above, but because most of these firms/schools/et al. usually confine their work to elections in their state, I separated them. The number in parenthesis is the score for each one's final state poll, and thus it represents its overall score.

Arizona

Myers/Grove AZ (6.8)
N. Ariz Univ: AZ (8.8)
ASU AZ (8.8)
Zimmerman AZ (12.8)

Colorado
Rocky Mountain/News 4 CO (6.4)

Florida
Datamar: FL (4.3)
Times Union/Sun Sentinel: FL (5.3)
St. Pete/Herald/News 9: FL (8.3)
Hamilton: FL (9.3)
Florida COC: FL (13.3)

Georgia
Democracy Corps: GA (9.4)

Indiana
Howey-Gauge: IN (6.9)
Selzer IN (7.9)

Michigan
Mitchell Research: MI (6.3)
EPIC-MRA: MI (8.3)
Detroit News: MI (10.3)

Minnesota
Star Tribune: MN (3.2)
Minnesota Public Radio: MN (8.8)
St. Cloud University: MN (19.2)

Montana
Montana St. Univ: MT (12.9)

New Hampshire
UNH: NH (4.1)

New Mexico
Albuquerque Journal: NM (6.7)

North Carolina
WSOC-TV: NC (5.4)
Civitas: NC (6.4)

North Dakota
MSU/The Forum: ND (10.6)
North Dakota UTU: ND (13)

Ohio
Univ. Cincinnati: OH (2)
Columbus Dispatch: OH (2)
Ohio Newspaper Poll: OH (3.4)

Pennsylvania

Muhlenberg: PA (4)
Franklin & Marshall: PA (6)
Temple Univ: PA (8)
Susquehanna: PA (11)

Virginia

CNU: VA (0.9)
Washington Post: VA (3.1)
VCU: VA (8.1)
Roanoke College: VA (12.1)

For the most part, a lot of these state polls did not do a great job in their final polls. There are standouts, of course: the Star Tribune in Minnesota, CNU and the Washington Post in Virginia, and all three pollsters in Ohio -- the University of Cincinnati, the Columbus Dispatch, and the Ohio Newspaper Poll -- all were very close to the final results in their states.