A lot of noise was made about the accuracy of polls and polling methods in the run-up to the election, so it's worth going back and taking a look at the polls to see how closely they match up with the actual results.
U.S. Senate
Lieberman defeated Lamont and Schlesinger 50%-40%-10%. See the full list of polls at RealClearPolitics.
Quinnipiac showed Lieberman leading Lamont and Schlesinger 50%-38%-8% (5% undecided) with a margin of error of 3.8% on November 6th. On November 1st, Quinnipiac showed Lieberman leading 49%-37%-8%. The margin of error was 3.2%. An earlier poll on October 23rd, however, showed Lieberman with a 17-point lead.
Rasmussen showed Lieberman leading Lamont and Schlesinger 48%-40%-9% on October 28th, with a 4.5% margin of error. On October 6th, Lieberman led 50%-40%-6%, with a 4.5% margin of error. American Research Group said on October 21st that Lieberman was ahead 49%-38%-8%. The poll had a margin of error of 4%. A Courant/UCONN poll out on October 11th showed Lieberman ahead 48%-40%.
There were a lot of others, including an outlier showing Lamont within 4% from Zogby and another from, well, Zogby showing Lieberman up 20%. But throughout October, Lieberman stayed in the high 40s, brushing 50% at times, while Lamont struggled to break 40%. Schlesinger picked up a few points following the debates, but never got to 10%. The trend that Lamont supporters were hoping for, in which Schlesinger took a significant amount of votes from Lieberman, never materialized. Undecideds broke for Schlesinger, perhaps, but there were so few of them that it hardly mattered. The polls were surprisingly accurate, especially as the election neared.
House Races
There was a lack of reliable polling information for all of the Congressional races--which seems to be par for the course.
The 2nd District wound up being the closest Congressional race in the country--and the polls showed it to a point. A Research 2000 survey (the JI-The Day poll) showed the race in a dead heat, with Courtney up by a single point (49%-48%) around Halloween. That poll is the most accurate. Reuters/Zogby had Simmons up by 5% in late October, while UCONN had Simmons up by 2% in mid-October. The polls could never make up their minds on this race--and neither could the voters, apparently.
The 4th District was a surprise win for Chris Shays, mostly because the polls showed him going down to defeat. Farrell led in the two latest polls by an increasingly large margin. Reuters/Zogby showed her winning by 7% in late October.
Perhaps the place where the polls got it wrong the most was in the 5th District, where Chris Murphy defeated Nancy Johnson by 12%. No poll forecast this. The closest any poll came to predicting the outcome of this one was a MajorityWatch (D) poll in October showing Murphy up by 8%. This poll was quickly dismissed as partisan. Republican pollsters had Johnson up by as much as 14% in September. Two independent polls were done at the end of October showing Murphy with a slight lead of 3-4%.
Governor's Race
Well, this one was easy. DeStefano took a Kennelly-esque whomping from Rell, losing to her 63%-35%. No poll quite dared to put her that far ahead, but all polls showed her up by a wide margin. Basically, all of the undecided voters listed in the polls shown broke for Rell.
So in summary, the polls did a lousy job of determining who would win in the state's Congressional races, but managed to predict with a decent degree of accuracy who would be our next Senator and Governor. This may be because Quinnipiac and Rasmussen, who seemed to be the most accurate, didn't poll Congressional races. Most of the numbers there came from partisan sources The Republican pollsters seemed to do particularly badly in predicting outcomes--if these are the numbers Karl Rove had, it's no wonder he was forecasting a Republican win.
Independent pollsters, with the exception of the obviously flawed Zogby, did pretty well, however.
9 comments:
Genghis,
Have you figured out any non-mind-bending ways to do a Senate race map?
I eagerly await your work.
FGMS,
The fact that Schlesinger didn't win a single town will make things easier. I might go the route of the primary maps and use shades of purple for Lieberman... either that or some sort of gnarly orangish color.
If you noticed, I did the easy maps first! :)
Looks like the Q poll had it dead on given the margin of error in the senate race. Recall the Lamont supporters calling Schwartz a lap dog for Lieberman because the results didn't come out in their favor. Come to find out, he was just reporting the poll data as he saw it, and Schwartz had it right.
I think the polls were about right. While the results may not be exactly the same as the polls, votes are the only way to exactly gauge the race on election day. Some people change their minds on election day and others don't, and that can't be measured accurately before the election.
I think some of the best polling was done in the 5th. The race was obviously breaking for Murphy, and it continued to do so until the polls closed. Were a poll done on election day before the polls closed, I think we would've seen a 10-14 point Murphy lead.
They'll be criticized until the end of time, but Quinnipiac does a hell of a job. They did the best work this fall.
My problem with Q-poll is simple, and even more egregious considering that there polling (with two outliers, one in the primary and one in the general, notwithstanding) turned out to be accurate - why did they stop releasing their party identification weights for this race? As Mystery Pollster (Mark Blumenthal) noted (I can't find the link - I know it exists, mea culpa), Q-poll has been very open in the past, but not in this race. Why? No clue - in retrospect it seems pretty clear that they were not doing anything sneaky. All the more reason why they should have been releasing their weighted party breakdowns!
Every televised story I saw on this race referenced the Q poll and ignored all others, and any poll held up consisently as the holy grail by the mass media can become a self-fulfilling prophecy; the problem is that when something like that happens, it's VERY difficult to prove. You can sit there all day and postulate whether Dean would have won the nomination in 2004 had the scream not been replayed a billion times, or whether Bush would have ever been elected had some of his worst gaffes been replayed by the media half as much, or whether Kerry would have won without the swiftboat ads, and not change a thing; media perception is not only paramount, but irreversable. Nonetheless, this issue is still not over by a long shot as far as many are concerned; there are many who still believe Schwartz did some push polling here and will continue to pursue him on the issue, and these are people who have nothing to do with the Democratic party or the Lamont campaign. For the sake of the process, I hope that if there really was some chicanery going on here, it will not go ignored simply because the push polling ended up being successful.
On another note, there WAS one poll not afraid to call the huge Rell lead. The Internet poll Pollmetrix released a last minute number for the Governor's race (at the same time they came out with their Senate race number), and it was 64% - 32%.
UConn was the clear loser in all this. They went a long time with no polling and then when they did they were really off. The pre and post debate polling was particularly questionable with the comparison between surveys that were in the field for different lengths in time. It also seemed like they changed methodology from past years. Anyone know if that is the case?
UConn was the clear loser in all this. They went a long time with no polling and then when they did they were really off. The pre and post debate polling was particularly questionable with the comparison between surveys that were in the field for different lengths in time. It also seemed like they changed methodology from past years. Anyone know if that is the case?
I didn't think "Majority Watch" was a partisan poll. Isn't the polling operation a joint effort between an "R" shop (RT Strategies) and a "D" shop (Constituent Dynamics) which they simply termed "majority watch"?
Anyway, they clearly got a lot of numbers right...
Post a Comment